Origins of Hot Jupiters
Rebekah I. Dawson1 and John Asher Johnson2
1Department of Astronomy & Astrophysics, Center for Exoplanets and Habitable
Worlds,The Pennsylvania State University, University Park, PA 16802;
email: [email protected] Department of Astronomy, Harvard University, Cambridge, MA 02138
Annu. Rev. Astron. Astrophys. . :1–50
https://doi.org/10.1146/((please add
article doi))
Copyright c© by Annual Reviews.
All rights reserved
Keywords
extrasolar planets, planet formation
Abstract
Hot Jupiters were the first exoplanets to be discovered around main
sequence stars and astonished us with their close-in orbits. They are a
prime example of how exoplanets have challenged our textbook, solar-
system inspired story of how planetary systems form and evolve. More
than twenty years after the discovery of the first hot Jupiter, there is
no consensus on their predominant origin channel. Three classes of hot
Jupiter creation hypotheses have been proposed: in situ formation, disk
migration, and high eccentricity tidal migration. Although no origin
channel alone satisfactorily explains all the evidence, two major ori-
gins channels together plausibly account for properties of hot Jupiters
themselves and their connections to other exoplanet populations.
1
arX
iv:1
801.
0611
7v1
[as
tro-
ph.E
P] 1
8 Ja
n 20
18
1. INTRODUCTION
Prior to the discovery of the first exoplanets, our conception of planetary systems was
generally informed by an observational sample of one: our own Solar System. In the origins
story informed by this limited yet well-studied sample, planets formed from a cloud of gas
and dust that collapsed into a disk. Beyond the snow line, where feeding zones are large and
solid icy materials are abundant, rocky cores grew quickly and accreted massive gaseous
atmospheres before the gas disk dissipated. In the inner disk, where feeding zones are small
and ices absent, rocky cores were too small to accrete gas and had to wait until after the
dissipation of the gas disk to grow by giant impacts. The planets maintained the orbits on
which they formed: circular and coplanar with rocky planets inside and gas giants outside.
The discovery of the first gas giants outside of the Solar System immediately upended
this Solar System-centric formation picture. The first exoplanet discovered around a main
sequence star, 51 Peg b (Mayor & Queloz 1995), has an astonishingly close-in orbit.1 Instead
of orbiting beyond several AU like our Solar System’s gas giants in the region we expected
giant planets to form, 51 Peg b orbits 10 times closer to its star than Mercury to the Sun.
51 Peg b belongs to the class of planets known as hot Jupiters, which we define here as gas
giants with masses greater than or equal to 0.25 Jupiter masses (0.83 Saturn masses) and
orbital periods shorter than 10 days (for comparison, Mercury’s orbital period is 88 days).
It was immediately clear from the discovery of 51 Peg b – and from the other hot Jupiters
that followed – that theories of planet formation needed revision. In parallel, the discovery
of the Kuiper belt has given us hints that even our own Solar System’s history was more
dramatic than we once assumed (e.g., Malhotra 1993, Thommes, Duncan & Levison 1999).
The discovery of the first hot Jupiters not only sparked a revolution in planet formation
theory but kick-started the field of exoplanet discovery and characterization. If the Solar
System paradigm for gas giants was the rule throughout the Galaxy, the progress of the field
of exoplanetary science would have been glacially slow. Radial velocity surveys require the
observation of at least one and ideally multiple orbits, which would take multiple decades
for a Jupiter analog. In contrast, early planet hunters could observe multiple hot Jupiter
orbits within a single week. Hot Jupiters also cause a larger reflex motion of the star than
a Jupiter analog, allowing observers to characterize the planets with fewer measurements
necessary to beat down the noise. Rapid discovery and confirmation of hot Jupiters led
to the build-up of a large sample of planets over a decade. The first statistical studies
of exoplanets, including the earliest measurements of the planet mass distribution with its
power-law rise toward less massive planets (Marcy & Butler 2000) and the planet-metallicity
correlation (Gonzalez 1997), were informed primarily by giant planets orbiting within 1 AU.
The rising discovery rate of exoplanets revealed that hot Jupiters, while playing a major
role in the field of exoplanets, are relatively rare in the Galaxy. Only 1% of Sun-like
stars host one, and the occurrence rate falls off around the most numerous stars, the M
dwarfs (Johnson et al. 2010). However, this rare class of planets has played an out-sized
role in our understanding of the internal structure, atmospheric composition, and orbital
architecture of giant planets outside the Solar System. These insights, like the sample
of planets discovered from Doppler surveys, are a direct benefit of the unexpected orbital
architecture of hot Jupiters. The probability of a transit scales with the inverse of a planet’s
1Although hot Jupiters surprised the modern astronomical community, their existence, discoveryvia radial velocity, and propensity to transit were proposed decades ago by Struve 1952.
2
In situ formation
Ex situ formation
Disk migration
Disk disappears
Eccentricity excitation
Disk disappears
Tidal migration
Figure 1 Three origins hypotheses for hot Jupiters: in situ formation (§2.1), disk migration
(§2.2), and tidal migration (§2.3).
semi-major axis, so a hot Jupiter at 0.05 AU is 100 times more likely to transit than a planet
at 5 AU. What was statistically impractical for a true Jupiter analog became a statistical
inevitability for the growing population of hot Jupiters discovered by radial velocity surveys.
Photometric monitoring of these hot Jupiters led to the first discovery of a transiting planet,
hot Jupiter HD 209458 b (Charbonneau et al. 2000, Henry et al. 2000). When a planet
transits, we can measure its size and combine the size and mass to derive a bulk density.
We can use secondary eclipses and transmission spectroscopy to study its atmosphere, a
step toward ultimately characterizing smaller, potentially habitable planets. We can also
measure the sky-projected stellar obliquity – the angle between the planet’s orbital axis and
the star’s spin axis – via the Rossiter-McLaughlin effect (Rossiter 1924, McLaughlin 1924).
The discovery of hot Jupiters paved the way for modern exoplanetary science by in-
spiring extensive theoretical work on several physical processes that were missing from our
pre-exoplanet story of how planetary systems form and evolve. In this review we provide
an overview of the various theories currently invoked to explain the origins of hot Jupiters
(§2). In §3, we summarize how properties of hot Jupiters themselves—including their ec-
centricities, host star obliquities, radii, semi-major axes, host star ages, and atmospheric
compositions—square with these theories. In §4, we synthesize tests of origin theories that
involve connecting hot Jupiters to other exoplanet populations. We conclude in §5 that
there are two formation channels that together are most consistent with observations, and
we outline future observational and theoretical studies that would clarify this picture.
2. OVERVIEW OF HOT JUPITER ORIGIN THEORIES
Here we provide a theoretical overview of the three main classes of hot Jupiter origin theory:
in situ formation (§2.1), disk migration (§2.2), and high eccentricity tidal migration (§2.3).
2.1. In situ formation
A major open question is whether hot Jupiters can form at their present day short orbital
periods. In situ formation is feasible if one or both of the two mechanisms proposed for
giant planet formation can operate close to the star: gravitational instability, in which part
of the proto-planetary disk fragments into bound clumps (e.g., Boss 1997; see Durisen et al.
2007 for a review), or core accretion, in which a rocky proto-planet core accretes many
www.annualreviews.org • 3
times its mass in gas from the proto-planetary disk (e.g., Perri & Cameron 1974, Pollack
et al. 1996; see Chabrier et al. 2014 for a review). Until recently, it was widely believed that
neither gravitational instability nor core accretion could operate at hot Jupiters’ close in
locations (Rafikov 2005, 2006) and hence hot Jupiters must have formed further from their
stars and migrated to their present-day orbits (§2.2–2.3). Here we review the feasibility of
in situ formation of hot Jupiters by either mechanism.
2.1.1. Gravitational instability: not plausible. A region of the proto-planetary disk will be
susceptible to gravitational instability if the free-fall time due to self gravity is sufficiently
rapid to overcome Keplerian sheer. The criterion for gravitational instability insta-
bility is parameterized as (Toomre 1964)
Q =2√kT/µ
GPΣgas= 130
(T
1500K
)1/2(2.3mH
µ
)1/2(3 day
P
)(2× 105 gcm−2
Σgas
). 1, 1.
where T is the temperature, µ is the mean molecular weight, mH is the mass of a hydrogen
atom, k is the Boltzmann constant, P is the orbital period, G is the universal gravitational
constant, and Σgas is the gas surface density. T ∼ 1500 K and Σgas ∼ 2 × 105 gcm−2 are
plausible disk conditions at P = 3 day.
It remains generally accepted that nebular conditions at hot Jupiters’ present day loca-
tions could not meet Q . 1 during the disk stage. At short orbital periods (denominator of
Eqn. 1), fast rotation supports the local gas against gravitational collapse . At high temper-
atures (numerator of Eqn. 1), thermal pressure supports the local gas against gravitational
collapse. For a disk heated by starlight, temperature increases as orbital period decreases
as T ∝ P−1/3, so temperatures are higher closer to the star. Although gas surface density
(denominator of Eqn. 1) is higher close to the star, typical gas surface density profiles are
not steep enough to compensate for hotter temperature and faster rotation close in the star.
For example in the minimum mass solar nebula, the gas surface density is constructed to
scale as Σgas ∝ P−1 (e.g., Hayashi 1981). In observed disks – at least at the wide separa-
tions we can observe – the density scales even more weakly with P (see Andrews 2015 Fig.
12, Section 5.2.2, references therein). Only a density profile steeper than Σgas ∝ P−7/6
would compensate for the fast rotation and high thermal pressure close to the star.
More importantly, even if the gas were dense enough for disk fragmentation, at short
orbital periods fragments would shear out before they could cool and contract. Following
Rafikov (2005) with an optical depth of unity (most efficient possible cooling) and adiabatic
index of 7/5, the criterion for fragments to cool before they are rotationally sheared is
ξ =5πkΣgas
σµPT 3= 2200
(Σgas
2× 105 gcm−2
)(3 day
P
)(1500K
T
)3
. 1, 2.
where σ is the Stefan-Boltzmann constant.
To meet both criteria simultaneously (gravitational instability, Q . 1, and cooling
before shearing, ξ . 1) requires implausibly high temperatures and gas surface densities:
T
1500K∼ 150
(3 day
P
)4/5(2.3mH
µ
)1/5
,
Σgas
2× 105 gcm−2∼ 1600
(3 day
P
)7/5(2.3mH
µ
)3/5
.
As argued by Rafikov (2005), such high temperatures would unbind the gas from the star.
Therefore we do not expect the conditions for gravitational instability close to the star.
4
2.1.2. Core accretion: requires huge build up of solids. The challenge for forming hot
Jupiters close to their stars via core accretion lies not in the accretion but in the core. In
the core accretion hypothesis (e.g., Pollack et al. 1996), giant planets form when a large,
solid ∼ 10 Earth mass core2 accretes gas from the nebula. Hot nebular gas can be accreted
by a massive core almost as easily as cool gas, because the accretion efficiency depends
primarily on the conditions at the radiative-convective boundary deep in the accreting
proto-planet’s atmosphere, not on the conditions in the nebula (e.g., Stevenson 1982, Lee,
Chiang & Ormel 2014, Piso et al. 2015). Throughout the proto-planetary disk, a giant
planet core, if formed, will accrete gas. The challenge lies in growing a sufficiently massive
core before the gas disk dissipates.
In gaseous proto-planetary disks, embryos grow by coagulation from planetesimals in
their reservoir of planet material. To grow into a core adequately large for gas accretion
(∼ 10M⊕; Rafikov 2006, Lee, Chiang & Ormel 2014, Piso, Youdin & Murray-Clay 2015),
the disk must satisfy two requirements (e.g., Pollack et al. 1996):
1. The timescale to grow the core is shorter than the gas disk lifetime, so that the gas
accretion stage can take place. Typical gas disk lifetimes are a few Myr (Fedele et al.
2010, Barenfeld et al. 2016).
2. The amount of mass in the feeding zone where the core grows is sufficiently large to
build up a ∼ 10M⊕ core.
Figure 2 displays these competing requirements. The first requirement is easier to meet
close to the star due to the short orbital timescales. We can estimate the coagulation
timescale of a 10 Earth mass core (Mcore = 10M⊕) as (Goldreich, Lithwick & Sari 2004,
Chiang & Laughlin 2013)
tcoag ∼ρ
2/3coreM
1/3core
ΣsolidsFP = 105yr
(ρcore
8gcm−3
)2/3(Mcore
10M⊕
)1/3(Σsolids
103gcm−2
)(P
3day
)F−1 3.
where F > 1 is the gravitational focusing factor, Σsolids is the solid surface density, and ρcore
is the bulk density of the core. Gravitational focusing and gas drag on planetesimals (e.g.,
Rafikov 2004; Chambers 2016 and references therein) can speed up this process further. In
Fig. 2, we use 10 Myr as generously long timescale for the dissipation of the gas disk.
Unfortunately for forming giant planets close to their stars, short orbital periods also
translate to tiny feeding zones and hence tiny core masses. Forming sufficiently massive
cores is more difficult close to the star, because
Mcore = 2πΣsolidsa∆RH . 4.
where a is the semi-major axis and the dimensionless quantity ∆ is the width of the feeding
zone in units Hill radii, RH, given by
RH = a
(Mcore
3M?
)1/3
, 5.
2Until recently, the existence of these cores had not been proven, even for Jupiter and Saturn(e.g., Guillot 2005). Using Juno gravity field measurements, Wahl et al. 2017 found a core mass of6–25 M⊕ for Jupiter. Hot Jupiter HAT-P-13b has a core mass of ∼ 11M⊕, Buhler et al. 2016.
www.annualreviews.org • 5
0.1
1.0
10.0
100.0
Cor
e m
ass
(MEa
rth)
10 gcm-2
20 gcm-2
40 gcm-2
Big enough core
Solid surface density at 1 AU (Σz,1):
0.1 1.0 10.0 100.0a (AU)
10-4
10-2
100
102
Cor
e gr
owth
tim
esca
le (M
yr)
Fast enoughgrowth
Figure 2 Intermediate semi-major axes facilitate conditions for giant planet formation in
high solid surface density disks. Top: Maximum core mass (set by the reservoir of planet
formation material, not disk lifetime; Eqn. 6) vs. semi-major axis. Bottom: core growth
timescale vs. semi-major axis (Eqn. 3). The solid surface density profile is Σsolids =
Σz,1(a
AU
)−3/2for three normalizations Σz,1 = 10, 20, 40gcm−2; the normalization is set to
increase by a factor of 4 at the ice line, 2.7 AU. The factor of 4 and ice distance of 2.7
AU are still under investigation (e.g., Lecar et al. 2006) and taken as illustrative values.
Under these assumptions, at small semi-major axes (i.e., where hot Jupiters are observed
today), the maximum core mass is too small to undergo runaway accretion, whereas at large
semi-major axes, the core cannot grow massive enough before the gas disk dissipates. When
the disk solid surface density is sufficiently high (red, blue), both conditions can be met at
intermediate semi-major axes (within the vertical dotted lines), but not if the solid surface
density is too low (purple). Values are illustrative to demonstrate the tension between the
core mass criterion and the core timescale criterion. The tension can potentially be resolved
by transporting solids to the inner disk, increasing the effective solid surface density and
hence the core mass, and by speeding up the core growth timescale at wider separations
through the mechanism of pebble accretion (see Johansen & Lambrechts 2017 for a review).
where M? is the host star mass. Eqn. 4 and 5 constitute a joint set of equations for Mcore
and RH. By solving for Mcore and substituting P for a using Kepler’s law, we obtain
Mcore = GΣ3/2solids
(M?
6π
)1/2
∆3/2P 2
= 0.005M⊕
(Σsolids
103gcm−2
)3/2(M?
M�
)1/2(∆
7
)3/2(P
3day
)2
. 6.
where ∆ = 7 is typical (e.g., Greenzweig & Lissauer 1990). The core mass is tiny, about half
6
a Moon mass. To grow a large core of ∼ 10M⊕ would require a factor of 160 enhancement
in Σsolids or ∆. Although the core can grow quickly near the star, its growth appears to
stall at a low mass. (See below for a discussion of how radial drift can greatly increase the
effective Σsolids by replenishing the planet formation reservoir.)
Recent studies have revisited the plausibility of forming hot Jupiters in situ (e.g., Lee,
Chiang & Ormel 2014, Lee & Chiang 2016, Batygin, Bodenheimer & Laughlin 2016, Boley,
Granados Contreras & Gladman 2016), motivated by the abundance of super-Earths, plan-
ets with masses between that of Earth and Neptune, at short orbital periods (Howard et al.
2010, Mayor et al. 2011). Many super-Earths are ∼> 10M⊕, large enough to be the cores
of giant planets. However, there are several important caveats to this connection. First,
super-Earths have a different orbital period distribution than giant planets, with their oc-
currence rate drops steeply with orbital period within 10 days (see Lee & Chiang 2017 and
references therein), as we will discuss further in §4.4. The relative occurrence rates can only
be reconciled if it much easier for cores grow in gas giant planets very close to the star than
further from the star, which is not what we expect from planet formation theory (§2.1).
Second, in situ formation tends to produce multiple nearby planets (e.g., Hansen & Murray
2013, Becker et al. 2015), as we will discuss further in §4.2. Finally, increasing ∆ beyond
7 requires mergers of multiple cores via giant impacts. However, damping from the gas
disk prevents orbits from crossing until after the gas disk dissipates. Lee & Chiang (2016)
highlight this argument as the strongest against forming hot Jupiters in situ. Furthermore,
∆ has a hard upper limit set by the escape velocity (e.g., Schlichting 2014), which limits
the self-stirring growth of cores’ eccentricities for orbit crossing. This results in a limit of
∆max .2vesc
RH2πP−1' 13
(P
3day
)1/3(ρ
8g cm−3
)1/6
. 7.
Therefore, given that we cannot make ∆ arbitrarily large, forming a ∼ 10M⊕ close to the
star can only be achieved by greatly enhancing the solid surface density beyond the minimum
mass solar nebula (e.g., Chiang & Laughlin 2013). Enhancing the entire disk mass (gas plus
solids) by two orders of magnitude would destabilize the disk (e.g., Schlichting 2014).
However, the effective Σsolids can be greatly increased by radial drift of solids. Highly
efficient transport and pile-up of solids in the inner disk could perhaps deliver 10 M⊕ of
solids to a tiny feeding zone. The mechanism of pebble accretion, in which cores grow
by accreting millimeter-to-centimeter-sized pebbles that drift from the outer disk, allows
cores to grow more quickly (see Johansen & Lambrechts 2017 for a review). However, core
growth stales when the core carves a gap, staunching the flow of pebbles, which build up at
a pressure bump. For a variety of assumptions about the disk properties, the limiting core
mass is < 1M⊕ at typical hot Jupiter separations. Therefore, the material for hot Jupiter
cores cannot come solely in the form of pebbles. A better understanding of how solids are
transported through the disk – as dust, pebbles, planetesimals, embryos, or proto-cores –
may resolve the open question of whether hot Jupiters can form in situ by core accretion.
2.2. Gas disk migration
In gas disk migration, torques from the gaseous proto-planetary disk can shrink a giant
planet’s semi-major axis from several AU to hundredths of an AU (e.g., Goldreich &
Tremaine 1980, Lin & Papaloizou 1986, Lin, Bodenheimer & Richardson 1996, Ida & Lin
2008; see Baruteau et al. 2014 for a comprehensive review). The planet exchanges angular
www.annualreviews.org • 7
momentum with the disk by perturbing nearby gas onto horseshoe orbits (via corotation
torques) and deflecting more distant gas (via Lindblad torques). The net Lindblad torque
tends to be inward (Ward 1997; see also Armitage 2013 for a pedagogical overview of migra-
tion torques). The strength and sign of corotation torques depend on the disk’s turbulent
viscosity, opacity, and radial entropy profile (e.g., Paardekooper & Mellema 2006, Duffell &
Chiang 2015); for disk conditions, corotation torques can drive a net outward migration. As
the planet deflects gas, it can clear a gap; the resulting reduction in its migration speed de-
pends on the disk viscosity and scale height. It was once believed that a sufficiently massive
planet could staunch the flow of gas across the gap and that the planet would move inward
on the timescale of disk’s viscous accretion onto the star (“type II migration”). Staunching
gas flow would have guaranteed inward migration at or above the disk’s viscous accretion
rate. However, recent simulations have shown that a significant amount of material passes
through the gap on horseshoe orbits (e.g., Duffell et al. 2014, Durmann & Kley 2015), and
the type II migration rate can be slower than viscous rate. The contribution of gas disk
migration to the origin of hot Jupiters remains ambiguous because the magnitude and sign
of the migration rate are highly sensitive to disk conditions.
When migration is faster than the disk lifetime, a giant planet risks migrating past the
hot Jupiter region and being tidally disrupted or engulfed by the star. When proposing disk
migration for the origin of the first discovered hot Jupiter 51 Peg b (Mayor & Queloz 1995),
Lin, Bodenheimer & Richardson (1996) invoked two mechanisms for halting 51 Peg b’s
migration: angular momentum transfer from the star and a “magnetocavity” in the inner
disk. In the former mechanism, an inward migrating hot Jupiter can sustain its close-in
orbit and avoid infall during the disk lifetime by extracting angular momentum from its star
via tides (Trilling et al. 1998). The planet may lose some mass through Roche lobe overflow
but remains giant. This tidal mass loss can contribute to reversing a hot Jupiter’s orbital
decay (e.g., Valsecchi et al. 2015). In the magnetocavity mechanism, the stellar magnetic
field creates a cavity in the innermost disk. Planet migration driven by the disk can shut
off when the hot Jupiter reaches the 2:1 resonance with the inner disk edge (e.g., Rice,
Armitage & Hogg 2008, Chang, Gu & Bodenheimer 2010). Another hypothesis is that hot
Jupiters halt at the dust sublimation radius (Kuchner & Lecar 2002). However, Eisner et al.
(2005) find that dust sublimation radii in observed disks are too large to be consistent with
hot Jupiters’ short orbital periods. In §3.4, we will review the consistency the semi-major
axis distribution with hot Jupiter origins hypotheses. If hot Jupiters arrive in the inner
disk via migration, the magnetospheric cavity, mass loss, and tidal interactions with the
star likely all play a role in setting its final location (e.g., Chang, Gu & Bodenheimer 2010).
Disk migration – at least in the single planet case – changes a planet’s semi-major
axis while keeping its eccentricity low, as depicted in Fig. 4. Planet-disk interactions can
excite eccentricities e under some disk conditions (Goldreich & Sari 2003, Tsang, Turner
& Cumming 2014, Duffell & Chiang 2015). However, Duffell & Chiang (2015) find the
excitation is limited to small values corresponding to epicyclic velocities approximately equal
to the sound speed within the disk. Above this maximum disk eccentricity excitation
value, the planet tends to undergo collisions with the gap edge that damp its eccentricity:
edisk .
√kT/µ
2πa/P= 0.015
(T
1500K
)1/2 ( 2.3mHµ
)1/2 (P
3day
)1/3
= 0.05(
T400K
)1/2 ( 2.3mHµ
)1/2 (P
600day
)1/3
. 8.
8
A migrating giant planet may capture planets into mean motion orbital resonances (e.g.,
Malhotra 1993, Lee & Peale 2002, Raymond, Mandell & Sigurdsson 2006). We summarize
what the properties of hot Jupiters’ companions reveal about their origins in §4.3.
2.3. High-eccentricity tidal migration
High-eccentricity tidal migration is another proposed mechanism to move a giant planet
from several AU to several hundredths of an AU. Turning a cold Jupiter into a hot Jupiter
requires reducing its orbital angular momentum by a factor of 10 and its orbital energy by a
factor of 100. Unlike disk migration (§2.2), in which the gravitational back reaction from the
disk changes the planet’s energy and angular momentum simultaneously, high eccentricity
migration can often be approximated as a two step process: reducing the planet’s orbital
angular momentum and then reducing its energy. (See §2.3.3 for exceptions.) During the
first step, a perturber extracts orbital angular momentum from the Jupiter by perturbing
the Jupiter onto a highly elliptical orbit. During the second step, the Jupiter tidally dissi-
pates its orbital energy through interactions with the central star. At periapse the Jupiter
undergoes close passages to its host star, which raise tides on the Jupiter. The Jupiter
dissipates energy as it stretches, changing shape to conform to the rapidly changing tidal
potential. During this tidal dissipation step, the hot Jupiter’s angular momentum is roughly
conserved as it circularizes to a final, close-in semi-major axis,
afinal = a(1− e2). 9.
For example, if the Jupiter begins the tidal dissipation stage at a = 4 AU (1 AU) and
circularizes to a = 0.04 AU, the perturber must have raised the Jupiter’s eccentricity to
e = 0.995 (0.98), corresponding to a periapse of 0.02 AU. Although these eccentricities are
quite high, they are plausible for eccentricity excitation mechanisms (§2.3.1) and in line
highly elliptical planets observed (e.g., HD 80606b has e=0.93, Naef et al. 2001; Fig. 4).
2.3.1. Eccentricity excitation: reducing the Jupiter’s angular momentum. Several mecha-
nisms have been proposed as sinks for Jupiter’s orbital angular momentum.
Planet-planet scattering: Planet-planet scattering converts Keplerian sheer (i.e.,
differences in angular velocity between planets with different semi-major axes) into angular
momentum deficit, triggering high eccentricity migration (e.g., Rasio & Ford 1996, Wei-
denschilling & Marzari 1996, Ford & Rasio 2006, Chatterjee et al. 2008). An example of
planet-planet scattering is shown in the top panel of Fig. 3. Planet-planet scattering can
take place in systems that form tightly packed (e.g., Juric & Tremaine 2008) or when high
eccentricities are generated by stellar fly-bys (e.g., Shara, Hurley & Mardling 2016). Al-
though planet-planet scattering also alters planets’ semi-major axes, it is highly unlikely
that a hot Jupiter can reach its present day short orbital period solely by scattering. Since
the system’s total orbital energy is conserved, a Jupiter would need to eject 100 planets
of its own mass to reduce its semi-major axis by 100. (The hot Jupiter can reduce its
semi-major axis through tidal dissipation in the next step.) In planet-planet scattering, the
planet’s eccentricity typically grows over a series of multiple close encounters. Eccentricity
growth via scattering is limited to an epicyclic velocity corresponding to the escape
velocity from the surface of the planet (e.g., Goldreich, Lithwick & Sari 2004, Ida, Lin &
www.annualreviews.org • 9
Nagasawa 2013, Petrovich, Tremaine & Rafikov 2014):
escatter .
√2GMp/Rp
2πa/P=(
Mp
MJup
)1/2 (RJup
Rp
)1/2 (P
50day
)1/3
= 0.2(
Mp
0.5MJup
)1/2 ( 2RJup
Rp
)1/2 (P
3day
)1/3
10.
Once the eccentricity exceeds escatter, the cross section for collisions exceeds the cross section
for scattering and planets merge rather than scatter during close encounters. At ∼AU
separations, the escatter exceeds 1 and Jupiters can be scattered onto highly elliptical orbits.
Secular interactions: Secular interactions are slow exchanges of angular momentum
between widely separated planets. Secular interactions take place over thousands or even
millions of years, depending on the separation and mass of the planets involved. Through
secular interactions, a Jupiter can deposit its angular momentum into other planets or
stars in the system on a timescale of many orbits. Planets can swap angular momentum
periodically (e.g., Petrovich 2015a) or chaotically (e.g., Wu & Lithwick 2011, Hamers et al.
2017), driving the Jupiter’s eccentricity to large values. The latter scenario requires three
or more planets and is known as secular chaos. (Chaos can occur in hierarchical two planet
systems as well, but requires large eccentricities and inclinations; see Wu & Lithwick 2011
for a discussion.) Kozai-Lidov cycles are a type of periodic angular momentum exchange
that also trade off mutual inclination and eccentricity (Kozai 1962, Lidov 1962, Naoz 2016).
Kozai-Lidov cycles can also occur in an initially coplanar system when the outer body is
on a highly elliptical orbit (Li et al. 2014b). Kozai-Lidov cycles driven by a star (e.g., Wu
& Murray 2003, Fabrycky & Tremaine 2007, Katz, Dong & Malhotra 2011, Naoz, Farr &
Rasio 2012, Petrovich 2015b) or planet (e.g., Naoz et al. 2011; see Teyssandier et al. 2013 for
distinctions between planet-planet and planet-star Kozai) have been widely hypothesized
to trigger high eccentricity tidal migration of hot Jupiters. In another scenario, multiple
secular frequencies within a system coincide, creating a secular resonance that elevates
planets’ eccentricities (e.g., Minton & Malhotra 2011, Xu & Lai 2016). Examples of several
types of secular excitation interactions are shown in Fig. 3.
Secular interactions conserve the system’s total orbital angular momentum. The proto-
hot Jupiter loses its angular momentum by taking a share of the system’s angular momen-
tum deficit, the difference between the system’s actual angular momentum and the angular
momentum it would have if all the bodies had circular, coplanar orbits. The angular mo-
mentum deficit could have originally been created from Keplerian sheer (i.e., via scattering
as discussed above). In the case of Kozai-Lidov cycles generated by a stellar perturber, a
widely separated binary may naturally form in a different plane than the planet-primary,
resulting in a mutual inclination that can drive Kozai-Lidov.
Timescales for exciting eccentricity: In planet-planet scattering, eccentricities are
excited on a synodic timescale at planets’ conjunctions. Each encounter is a random kick,
and the eccentricity grows as a random walk. In Figure 3, first row, the periapse shrinks to
a value sufficient for tidal circularization on a timescale of thousands of years.
Secular excitation of eccentricities occur on secular timescales of thousands of orbits.
The timescale is primarily set by the perturber’s mass and orbit. Rows 2–4 of Fig. 3 feature
eccentricity excitation by secular cycles. In row 3, a 1 M� Jupiter at 1000 AU causes ∼ 20
Myr Kozai-Lidov eccentricity oscillations of a mutually inclined coplanar Jupiter at 5 AU
(Fabrycky & Tremaine 2007). In row 4, a coplanar, 3.3 MJup Jupiter at 8 AU causes ∼ 1
Myr eccentricity oscillations of a coplanar Jupiter at 1 AU (Petrovich 2015a). In row 2, a
10
0.002 0.004 0.006 0.008 0.010
0.01
0.10
1.00
10.00P-
P sc
atte
ring
BN120.1 1.0
decoupling
semimajor axisperiapse
20 40 60 80 1000.01
0.10
1.00
P-P
Koza
i
N+11
110 111 112 113
0 20 40 60 80
0.1
1.0
P-* K
ozai
FT071000 2000 3000 4000 5000
1 2 3 4 5
0.1
1.0
P-P
Sec
Cop
l
P15
10 100
293.0 293.2 293.4 293.6 293.8 294.00.01
0.10
1.00
Sec
Cha
os
WL11
294.0 295.7 297.4 299.1 300.8 302.5t (Myr)
dist
ance
(AU
)
Figure 3 Dynamical histories leading to hot Jupiter: planet-planet scattering (Beauge &
Nesvorny 2012), planet-planet Kozai (Naoz et al. 2011), planet-star Kozai (Fabrycky &
Tremaine 2007), planet-planet coplanar secular (Petrovich 2015a), and secular chaos (Wu
& Lithwick 2011). The planet loses angular momentum/is perturbed to a small periapse
(dotted line), i.e., high eccentricity. When the planet gets close enough to the star (red),
it tidally circularizes, i.e., loses orbital energy, and its semi-major axis (solid line) shrinks.
Eventually the planet decouples from its perturber (blue vertical dashed line). Simulation
data was extracted from each cited article. See text for a discussion of timescales.www.annualreviews.org • 11
3 MJup Jupiter at 61 AU causes eccentricity oscillations of a mutually inclined Jupiter at 6
AU on two timescales: a shorter Kozai-Lidov timescale of ∼ 1 Myr and longer envelope of
∼ 20 Myr caused by the perturber’s eccentricity (Naoz et al. 2011). See Li et al. (2014b),
Antognini (2015) for Kozai-Lidov eccentricity excitation timescales in the regime where the
perturber is nearby and/or eccentric. In secular chaos, the proto-hot Jupiter’s eccentricity
diffusively randomly walks over many secular timescales. In the example shown in row 5
of Fig. 3, the eccentricity reaches a value large enough for tidal circularization after 300
Myr (Wu & Lithwick 2011). Timescales can range from millions of years to many billions
of years (e.g., the Solar System, Laskar 2008).
2.3.2. Tidal dissipation: reducing the hot Jupiter’s orbital energy. Once the Jupiter’s orbit
is sufficiently elliptical, tidal dissipation in the planet shrinks and circularizes its orbit. Tidal
dissipation in the planet decreases the planet’s orbital energy but keeps its angular momen-
tum constant. Once the hot Jupiter is gravitationally decoupled from the perturber(s) that
originally removed the hot Jupiter’s angular momentum, its orbit evolves according to
afinal = a(t)[1− e(t)2]. 11.
where a(t) and e(t) are the time evolution of the semi-major axis and eccentricity respec-
tively. The link between a and e lets us rewind a hot Jupiter’s tidal evolution history. For
example, a hot Jupiter observed today on a low eccentricity orbit at 0.04 AU that under-
went tidal migration had e = 0.9 at a = 0.2 AU and e = 0.99 at 2 AU. Tidal evolution
tracks of constant angular momentum are included in Fig. 4.
Regardless of the details of the tidal physics, the tidal evolution timescale has a strong
dependence on afinal. For example, in the constant tidal time lag model (e.g., Eggleton,
Kiseleva & Hut 1998)
a/a ∝ a8final. 12.
When the initial orbit at the beginning of tidal circularization is highly elliptical, the initial
periapse a(t = 0)[1 − e(t = 0)] ≈ 12afinal because tracks of constant angular momentum
are defined by afinal = a(1 − e2) and a(1 − e2) ≈ 2a(1 − e) for e → 1. The tidal evolution
timescale is highly sensitive to the initial periapse and spans many orders of magnitude. A
planet that takes 1 Myr to circularize to afinal = 0.03 AU (initial periapse 0.015 AU) would
take longer than the age of the universe to circularize to afinal = 0.1 AU (initial periapse
0.05 AU). Figure 3 features tidal circularization on a range of timescales from ∼ 0.1 Myr
(row 2) to Gyrs (row 3).
It is important to remember that high eccentricity tidal migration is caused by tides
raised on the planet by the star, not tides raised on the star by the planet. Tides raised
on the star by the planet are typically much less efficient and much more sensitive to a
than to afinal. If and when a hot Jupiter arrives at a small afinal, its tides raised on the
star may drive orbital evolution on a longer timescale (e.g., Valsecchi et al. 2015). If the
planet arrives via tidal dissipation in the planet at an orbital period longer (shorter) than
the star’s rotational period, its orbit will expand (shrink) during this final stage driven by
tidal dissipation in the star.
2.3.3. Decoupling from the perturber. The hot Jupiter’s tidal interactions with the central
star eventually decouple it from the gravitational perturbations that originally raised its
eccentricity. In the case of planet-planet scattering, the planet decouples when its orbit
12
shrinks sufficiently to avoid subsequent close encounters with its scatterer (Fig. 3, top
panel). In the case of secular interactions, the planet decouples as tides shrink its semi-
major axis and precession from general relativity or tides exceeds the precession caused by
the secular perturber. Figure 3 features a range of decoupling timescales. In the top panel,
we see a quick decoupling triggered by a kick and subsequent tidal circularization. In panel
3, we see a gradual decoupling as the semi-major axis shrinks and eccentricity oscillations
are gradually quenched.
Some binary companions are too weak to even compete with general relativity in the
first place and high eccentricity migration is never triggered, as we will discuss further in
§4.3. In order for a perturber companion to excite the eccentricity of the proto-hot Jupiter,
the precession caused by the perturber must dominate over the precession caused by the
central star, including due to general relativity , the tidal distortion of the planet, and the
oblateness of the star (e.g., Fabrycky 2010, Eqn. 15–17). Dong, Katz & Socrates (2014)
derived quenching criterion (their Eqn. 4). In the limiting case of initially circular orbits
for the proto-hot Jupiter and perturber and a proto-hot Jupiter with an orbital period much
larger than a hot Jupiter’s, this criterion is approximately(mperturber
0.1M�
)(800yr
Pperturber
)2
>
(M?
M�
)5/3(1yr
P
)7/3(3day
PHJ
)1/3
13.
Figure 3 depicts examples of decoupling in cases of planet-planet scattering and secular
interactions. In some cases – particularly when the perturber is a massive, nearby planet
(Dong, Katz & Socrates 2014) – a proto hot Jupiter only decouples at the end of its tidal
evolution and can undergo eccentricity oscillations throughout its journey (e.g., Fig. 3,
row 3). In that case, Eqn. 11 does not strictly apply: the planet’s angular momentum
is not constant but changing due to continued interactions with the perturber. However,
the angular momentum averaged over an eccentricity cycle is roughly conserved throughout
tidal evolution (e.g., as argued by Socrates et al. 2012).
3. TESTING HOT JUPITER ORIGIN THEORIES USING PROPERTIES ANDCORRELATIONS
The observed properties of hot Jupiters and their host stars—and correlations among those
properties—can be used to test theories of the origins of hot Jupiters. In this section we
summarize how hot Jupiter properties square with their origin via in situ formation (§2.1),
disk migration (§2.2), and different varieties of high eccentricity tidal migration (§2.3). We
divide the properties into intrinsic properties (this section) and connections of hot Jupiters
to other populations (§4).
3.1. Eccentricities of hot Jupiters
The present-day eccentricities of giant planets are relics of their origins. Each orbital evolu-
tion process we described in §2—including planet-disk interactions, planet-planet scattering,
secular interactions, and high eccentricity tidal migration—plays out differently in the pa-
rameter space of eccentricity and semi-major axis. We illustrate how different processes
populate the (semi-major axis, eccentricity) space and compare to the observed population
in Fig. 4. In this subsection, we focus on the eccentricities of hot Jupiters themselves
(P < 10 days in Fig. 4). (We will assess the predictions of hot Jupiter origin theories
www.annualreviews.org • 13
for longer period giant planets eccentricities in §4.) A hot Jupiter that gets close to its
host star can collide with the star (Fig. 4, dark yellow region), be tidally disrupted (light
yellow region), or tidally circularize (red region). In the observed distribution (Fig. 4), hot
Jupiters are absent from the stellar collision and tidal disruption regions.
Out to ∼3 day orbital periods, most hot Jupiter orbits are consistent with circular.
Jupiters at the shortest orbital periods likely have such fast tidal circularization timescales
(Eqn 12) that we are unlikely to catch them on elliptical orbits. The circularization limit for
hot Jupiters can be used to calibrate theories of tidal dissipation inside planets (e.g., Hansen
2010, Socrates et al. 2012). An outstanding issues is that the tidal dissipation required to
circularize these hot Jupiters is stronger than expected. Socrates et al. (2012) found that
the tidal efficiency of hot Jupiters must be at least 10 times higher than the tidal efficiency
of our own Solar System’s Jupiter. Current tidal models of dissipation by inertial waves do
not predict such strong dissipation. Other processes – such as the elliptical instability (e.g.,
Barker 2016) – may enhance the dissipation efficiency.
14
ecce
ntric
ity
semi-major axis (AU)
0.45
0.63
0.77
0.89
1.00
0.01 0.10.00
1.0
planet-disk interactions
planet-planet scattering
Stellar collis
ion
Tidal
disru
ption
Higheccentricitymigration
Tidal
migr
ation
0.01 10.0semi-major axis (AU)
ecce
ntric
ity
0.00
0.45
0.63
0.77
0.89
1.00
0.01 0.10 1.00
Fe-richFe-poor
0.51.01.5
rate
(%)
IncompleteRV observedC+08
1 3 10 100 1000Period (days)
www.annualreviews.org • 15
Figure 4 Regions of parameter space (eccentricity vs. semi-major axis) populated by differ-
ent mechanisms of orbital evolution. Note that the y-axis is linear in 1 − e2. Eccentricity
and semi-major refer to a planet’s instantaneous orbit: planets move through the diagram
as their orbits evolve. For example, a planet could form at (a =5 AU, e = 0), have its
eccentricity excited by secular interactions with a companion and enter the red region, and
tidally circularize a short orbital period. Top: Schematic. Middle: Observed radial-velocity
planets with masses M sin i > 0.3MJup, where i is the sky-plane orbital inclination (90◦
is edge on). Planets disturbed into the yellow region would be tidally disrupted. Planets
in the red region have small enough angular momentum for tidal circularization and those
in the red dashed-outlined striped region could have started at high eccentricities (large
semi-major axes) without having been tidally disrupted. Planet-planet scattering excites
eccentricities to values limited by the planets’ capacity to scatter rather than merge (gray
region). Planet-disk interactions can alter semi-major axes and damp or modestly excite
eccentricities (blue region). The solid red, gray, and blue arrows are examples of orbit
changes via high eccentricity migration, eccentricity excitation, and planet-disk scattering
respectively. Illustrative parameters are chosen to delineate the approximate region for each
mechanism; for example, the slope of the gray region depends on the planet’s mass. See §4.3
for a discussion of if and how these mechanisms can populate the unshaded white region of
eccentric warm Jupiters. Observed giant planets are color-coded by host star metallicity:
those greater than or equal to solar within one sigma are red and others are blue. Bottom:
histogram of semi-major axes of observed radial-velocity planets and occurrence rates in-
ferred by Cumming et al. (2008) (C+08). Planets compiled from exoplanets.org (Wright
et al. 2012) and exoplanets.eu (Schneider et al. 2011). Eccentricities and upper limits
taken from Bonomo et al. (2017) where available. The histogram is computed according to
semi-major axis; the orbital period axis is for a Sun-like star. The observed histogram is
normalized to total to 7.5% for planets between 0.03–3 AU and 0.3–10 Jupiter masses (see
Dong & Dawson 2016 5.1 discussion of Cumming et al. 2008 results).
16
In the 3–10 day orbital period range, some hot Jupiters occupy moderately elliptical
orbits (0.2 < e < 0.6). If they originate through in situ formation (§2.1) or disk migration
(§2.2), their eccentricities would need to be excited at their present day close-in locations.
However, as discussed in §2, interactions with the gas disk (Eqn. 8) and planet-planet scat-
tering in situ (Eqn. 10) cannot account for the moderate eccentricities observed. Possibly
an outer planet could secularly force the eccentricity, but the perturber would need to be
massive and/or nearby to overcome precession from general-relativity and tides. In the
limit of Laplace-Lagrange secular excitation (which applies for low-to-moderate eccentrici-
ties and inclinations), an approximate criterion for the precession caused by the perturber
(e.g., Fabrycky 2010, Eqn. 21) to exceed general-relativistic precession is(mperturber
MJup
)(37day
Pperturber
)2
&
(M�
M?
)5/3(3day
P
)8/3
14.
As we will discuss further in §4.2, such companions have been ruled out for many moder-
ately eccentric hot Jupiters. Similar criteria can be derived for the perturber precession to
overcome the precession caused by the tidal distortion of the planet or rotation of the star.
In contrast, high eccentricity tidal migration (§2.3) provides a natural explanation for
these hot Jupiters on moderately eccentric orbits. According to this theory, moderately
elliptical hot Jupiters are in the process of tidal circularization. However, the occurrence of
these moderately eccentric Jupiters may not be compatible with all tidal migration theories.
Petrovich (2015b) found that high eccentricity tidal migration spurred by stellar binary
Kozai-Lidov cannot account for the number of moderately eccentric Jupiters relative to hot
Jupiters: they predict only 1 moderately eccentric hot Jupiter for every 300 hot Jupiters
on circular orbits. Among the 228 currently known hot Jupiters (m > 0.1MJup) in the
3–10 day orbital period range, 31 have eccentricities constrained at two sigma to be below
e < 0.1 (i.e., that could have been excited in situ), 10 have eccentricities constrained at
two sigma to be e > 0.2 (i.e., that most likely could not have been excited in situ and
are evidence of high eccentricity tidal migration), and the rest have intermediate or poorly
constrained eccentricities. Because the tidal circularization timescale is shorter for more
less planets among planets of similar radii, less massive planets should be circularized out
to larger host star separations (Pont et al. 2011). In a large sample of hot Jupiters with
uniformly derived eccentricities and upper limits, Bonomo et al. (2017) found that this
trend is present, supporting high eccentricity migration.
Why do we observe both circular and eccentric hot Jupiters at the same orbital periods?
Under the high eccentricity migration hypothesis, circular hot Jupiters with 3–10 day orbital
periods have completed their tidal circularization. Possibly circular hot Jupiters began their
migration earlier than eccentric hot Jupiters at similar orbital periods (a possibility we will
return to in §3.5) or have more efficient tidal dissipation properties. Another possibility
is that multiple hot Jupiter formation channels are at work and some low eccentricity
hot Jupiters originated via disk migration or in situ formation. Dawson & Murray-Clay
(2013) suggested two formation channels for hot Jupiters—one of which is high eccentricity
tidal migration—based on trends with host star metallicity. Eccentric hot Jupiters (red
diamonds, Fig. 4) orbit metal rich stars, while circular hot Jupiters orbit both metal
rich and metal poor stars. (See also Shabram et al. 2016). The correlation with host
star metallicity may indicate that high eccentricity migration was spurred by some type of
planet-planet gravitational interaction (§2.3), as giant planet occurrence is correlated with
host star metallicity (Gonzalez 1997, Santos, Israelian & Mayor 2001, Santos et al. 2003,
www.annualreviews.org • 17
Table 1. Evidence for origins hypotheses of hot Jupiters (HJ), including links to warm
Jupiters (WJ)
Evidence In situ formation Disk migration Tidal migration(§2.1) (§2.2) (§2.3)
Elliptical HJ (§3.1) X X XHJ obliquitities (§3.2) O O OInflated HJ radii (§3.3) X X XHJ semimajor axes (§3.4) X X XT Tauri HJs (§3.5) X X XHost star ages (§3.5) O O OAtmospheres (§3.6) T T TOccurrence rates (§4.1) T X XCompanions (§4.2) X X XHJ vs. WJ occurrence (§4.3) X X XCircular WJs (§4.3) X X XElliptical WJs (§4.3) X X XNearby WJ companions(§4.3)
X T X
Small planets (§4.4) X X THoptunes (§4.4) X T X
Note. — X: consistent, X: inconsistent, T: no clear prediction from theory yet, O:additional or complementary observations needed
Santos, Israelian & Mayor 2004, Fischer & Valenti 2005, Sousa et al. 2011).
In summary, the existence of moderately eccentric hot Jupiters is evidence at least a
fraction hot Jupiters underwent high eccentricity tidal migration. Furthermore, although
we do not know exactly where they began their tidal circularization, they must have orig-
inated far enough from the star for planet-planet scattering and/or secular excitation to
be effective. We will return to the possibility that multiple origins channels contribute
substantially to the hot Jupiter population throughout the review. We will tabulate which
properties of hot Jupiters the three origins hypotheses explain or fail to explain in Table 1.
3.2. Obliquities of hot Jupiters’ host stars
Hot Jupiter host star obliquities were once pursued as the Rosetta stone of hot Jupiters’ ori-
gins. The host star obliquity refers to the angle between the star’s spin angular momentum
vector and the hot Jupiter’s orbital angular momentum vector. The stellar obliquity can be
measured in projection for individual stars using the Rossiter-McLaughlin effect (McLaugh-
lin 1924, Rossiter 1924; see Triaud 2017 for a review), the rotational broadening of stellar
spectral lines (e.g., Schlaufman 2010), star spot crossings (e.g., Sanchis-Ojeda et al. 2011),
Doppler tomography (e.g., Collier Cameron et al. (2010)), gravity darkening (e.g., Barnes
2009), or asteroseismology (e.g., Huber et al. 2013) and for an ensemble based on star spot
modulation amplitudes (e.g., Mazeh et al. 2015, Li & Winn 2016). Many hot Jupiters ap-
18
pear well-aligned with their host stars’ spin, but others are dramatically misaligned, even
polar and retrograde (e.g., Albrecht et al. 2012 and references therein).
We might naively expect planets that form in an accretion disk to be on orbits aligned
with the star’s rotational plane. In this simple picture, hot Jupiters that originate in situ
(§2.1) or via disk migration (§2.2) would maintain aligned orbits (even as they interact
gravitationally with the gas disk, e.g., Bitsch et al. 2013). In contrast, the gravitation
interactions that reduce the magnitude of the Jupiter’s orbital angular momentum— trig-
gering high eccentricity tidal migration (§2.3)—would commonly change the direction of
the angular momentum vector. Different mechanisms for generating the Jupiter’s eccen-
tricity (planet-planet scattering, Kozai-Lidov cycles driven by a binary, etc.) would result
in different distributions of stellar obliquities (e.g., Fabrycky & Tremaine 2007, Chatterjee
et al. 2008, Naoz et al. 2011, Teyssandier et al. 2013, Li et al. 2014a). The theoretically
predicted obliquity distribution could be compared to the observed distribution to identify
the predominant eccentricity generating mechanism and even tease out the contributions of
multiple pathways for hot Jupiter origins (e.g., Fabrycky & Winn 2009, Morton & Johnson
2011, Naoz, Farr & Rasio 2012). Tides raised on the planet, which erase hot Jupiters’ eccen-
tricities (§3.1), leave stellar obliquities intact. The hot Jupiter’s orbital angular momentum
is comparable to a star’s spin angular momentum and timescale for realigning the star is (in
this simple picture) comparable to tidal decay of the hot Jupiter. Although tides raised by
the planet on the star could erase the stellar obliquity, this process would require complete
tidal decay and disruption of the hot Jupiter (e.g., Hut 1981). In this naive picture, stellar
obliquities would be a powerful, unambiguous indicator of hot Jupiters’ origin channel.
Unfortunately, interpreting stellar obliquities has proved to be far more challenging in
reality. The following complications weaken our expectations that in situ formation and
disk migration result in low obliquitities and that high eccentricity tidal migration should
go hand in hand with spin-orbit misalignments:
• Despite expectations to the contrary, spin-orbit alignments may be erased by tides
raised on the star by the planet. As the sample of hot Jupiter host star obliquities
grew, correlations between these obliquities and stars’ tidal properties raised the alarm
that the observed obliquity distribution, rather than purely reflecting the hot Jupiters’
origins, has been sculpted by tidal realignment. Schlaufman (2010) and Winn et al.
(2010) found that among hot Jupiter hosts, only hot stars’ spins are misaligned with
their hot Jupiters’ orbits; cool stars may be more easily tidally realigned (Winn
et al. 2010, Albrecht et al. 2012). The temperature cut-off coincides with the Kraft
break (Kraft 1967), implicating stellar spin-down in the realignment process (e.g.,
Dawson 2014). Empirical evidence has emerged that planets can indeed influence
stars’ rotation (e.g., Catala et al. 2007, Kovacs et al. 2014, Poppenhaeger & Wolk
2014, Miller et al. 2015). However, the interpretation that tidal realignment has erased
primordial obliquities has a major unresolved theoretical problem: how can a mere
planet realign an entire star’s spin without the planet sacrificing all of its angular
momentum (and without retrograde planets ending up perfectly anti-aligned; e.g.,
Lai 2012, Damiani & Lanza 2015, Lin & Ogilvie 2017)?
• Hot Jupiters originating via in situ formation or disk migration may be misaligned.
Mechanisms have recently been proposed for misaligning the disk or star itself, in-
cluding a binary perturber (e.g., Batygin 2012, Spalding & Batygin 2015) or stellar
fly-by (e.g., Xiang-Gruess 2016) tilting the disk; the star misaligning itself through
www.annualreviews.org • 19
Ex situ formation
Disk disappears
Misalignment by orbital perturbation
Disk disappears
In situ formation
Disk migration
Tidal migration stellar tidal
realignment
misalignment by internal gravity
wavesInitial disk misalignment
Figure 5 Misalignments between a hot Jupiter’s orbital angular momentum vector (purple
dashed) and its host star’s spin axis (yellow dotted) are influenced by a variety of physical
processes including primordial misalignment of the disk the planet forms from, misalignment
of the planet’s orbit by a perturber, and realignment via tides raised on the star. In the
absence of primordial misalignment, the star spin (dashed yellow) is initially taken to be
perpendicular to the disk. Additional misalignment pathways not shown include chaotic
evolution of the host star’s spin axis during tidal evolution with Kozai-Lidov cycles (Storch,
Anderson & Lai 2014) and misalignment by injection of an earlier planet (e.g., Matsakos &
Konigl 2015).
internal gravity waves (e.g., Rogers, Lin & Lau 2012); change in the spin of the proto-
stellar cloud over the star formation timescale (e.g., Fielding et al. 2015); injection
of an earlier planet (e.g., Matsakos & Konigl 2015); or chaotic evolution of the host
star’s spin axis during tidal evolution with Kozai-Lidov cycles (Storch, Anderson &
Lai 2014) . Some of these mechanisms can plausibly account for the observed stellar
temperature dependence without the planet needing to tidally realign the star. For
example, Rogers, Lin & Lau (2012) found that misalignment through internal gravity
waves only affects hot stars above the Kraft break. Spalding & Batygin (2015) found
that magnetic torques can realign the stellar spin-axes of lower-mass stars, which tend
to be cooler, with the disk plane.
Figure 5 schematically depicts how the above complications affect our interpretation of hot
Jupiters’ host star obliquities.
The case remains open on whether hot Jupiters’ host star obliquities implicate a par-
ticular origin scenario, require multiple origins scenarios, or primarily reflected physical
processes unrelated to hot Jupiters’ origins. A promising path is to compare hot Jupiters’
host star obliquities to obliquities of stars hosting other classes of planets: we will discuss
progress and prospects for comparative studies of obliquities in §4.2.
3.3. Hot Jupiter radius inflation
A number of hot Jupiters—including HD 209458b, the first hot Jupiter to be discovered
to transit (Charbonneau et al. 2000, Henry et al. 2000)—have radii larger than expected
from internal structure models. Hot Jupiters’ inflated radii require an additional heat
source, such as tidal heating during high eccentricity migration (e.g., Bodenheimer, Lin &
20
Mardling 2001), thermal tides caused by stellar irradiation (e.g., Arras & Socrates 2010,
Socrates 2013), or deposition of stellar irradiation energy into the interior (e.g., Guillot
& Showman 2002, Batygin & Stevenson 2010, Youdin & Mitchell 2010). All three classes
of hot Jupiter origins hypotheses can be consistent with our current understanding of hot
Jupiter inflation but the history of inflation plays out differently in each. In each case, the
hot Jupiter begins its life hot and inflated from formation (e.g., Spiegel & Burrows 2012)
and heat mechanisms must sustain inflation or re-inflate the hot Jupiter. Hot Jupiter radii
are observed to be strongly correlated with stellar flux (e.g., Weiss et al. 2013), supporting
the interpretation that one or more stellar irradiation deposition mechanisms plays a role.
The scenarios of in situ formation and disk migration only require a mechanism to
sustain inflation. Disk migration—which by definition occurs early in the planet’s lifetime—
naturally delivers the hot Jupiter close to its star before it cools on a ∼ 10 Myr timescale
(e.g., Spiegel & Burrows 2012, Wu & Lithwick 2013). One or more stellar irradiation
mechanisms can subsequently keep the hot Jupiter inflated as it ages.
Tidal migration can also be compatible with inflated hot Jupiter radii. One possibility
is that tidal heating re-inflates the proto-hot Jupiter after it cools (e.g., Ibgui, Spiegel &
Burrows 2011): in this case, the hot Jupiter does not need to arrive early in the star’s
lifetime. Tidal heating shuts off as the orbit nears circularization (e.g., Leconte et al. 2010,
Hansen 2010), so stellar irradiation mechanisms could take over at that point. A second
possibility is that thermal tides or stellar irradiation deposition mechanisms may be capable
of re-inflating a hot Jupiter after its arrival (e.g., Hartman et al. 2016, Lopez & Fortney
2016). An important caveat is that heating mechanisms that decay with depth, such as
the commonly invoked stellar irradiation deposition mechanism of ohmic dissipation (e.g.,
Batygin & Stevenson 2010), have extremely long re-inflation timescales and are unable to
re-inflate a Jupiter within the star’s lifetime if it cools before arrival (e.g., Ginzburg & Sari
2016; see also Wu & Lithwick 2013).
A final possibility is that the inflated hot Jupiters are those that tidally migrated early
in the star’s lifetime, before cooling. This possibility requires hot Jupiters to begin high
eccentricity tidal migration at a young age, which is plausible because the dissipation of
the gas disk may naturally trigger eccentricity excitation. When the gas dissipates, planets
cushioned by the gas disk begin to scatter and secular excitation turns on as the precession
from the gas disk shuts off. Hot Jupiters with smaller afinal migrate more quickly (Eqn.
12), so this scenario is consistent with inflated hot Jupiters being closer to their stars (i.e.,
as argued by Wu, Murray & Ramsahai 2007).
3.4. Hot Jupiter semi-major axis distribution
The three origins channels in §2 each make different predictions for the distribution of hot
Jupiter semi-major axes observed today (Fig. 4). The small end of the semi-major axis
distribution is affected by tidal disruption and large end by where formation and migration
can effectively deposit hot Jupiters. A modest peak in the distribution occurs at ∼ 3 days,
a feature known as the three-day pile-up. We caution that this feature appears misleadingly
large in plots that include hot Jupiters discovered by ground-based transit surveys, for which
selection effects sculpt a prominent three-day pile-up (see Gaudi, Seager & Mallen-Ornelas
2005 for a discussion of these selection effects). Because the statistical significance of the
three-day pile up has not been definitively established (i.e., whether it is a pile-up or simply
a drop off interior to 3 days), here we focus on the inner edge of the hot Jupiter region.
www.annualreviews.org • 21
Hot Jupiters’ present day orbits are consistent with tidal disruption limits (yellow re-
gions, Fig. 4). A hot Jupiter will be tidally disrupted inside the Roche limit, aRoche,
aRoche ' fpRp
(M?
Mp
)1/3
,
PRoche '2πf
3/2p R
3/2p
G1/2M1/2p
= f3/2p
(RpAU
)3/2(M�
Mp
)1/2
= 0.79 days
(fp2.7
)3/2(Rp
1.3RJup
)3/2(MJup
Mp
)1/2
, 15.
where Rp is the planet radius, Mp is the planet mass, M? is the stellar mass, and fp is a
dimensionless scale that depends on physical properties of the body. See the introduction
of Faber, Rasio & Willems (2005) for a pedagogical review of the Roche limit. The Roche
limit is related to the Hill radius (Eqn. 5): RH = Rp when a = aRoche for fp = 31/3 ≈ 1.44.
The limit fp = 31/3 corresponds to the maximum distance at which a test particle on the
surface of a perfectly spherical planet can remain at rest on the surface. For a planet subject
to tidal disruption, fp depends on the material properties of the Jupiter. From three-
dimensional hydrodynamical simulations of tidal disruption of giant planets, Guillochon,
Ramirez-Ruiz & Lin (2011) found that fp ≥ 2.7. In their simulations, planets at fp ∼ 2.7
are not immediately tidally disrupted but destroyed after mass loss and re-accretion over a
number of close encounters. The re-accretion makes the planet puffier and easier to destroy
in subsequent encounters. In Fig. 6, we plot a/aRocheof observed hot Jupiters (setting
fp = 2.7). All hot Jupiters have a/aRoche > 1 today. (WASP-19b has a/aRoche slightly
less than 1 but consistent within the uncertainties.) However, some of these Jupiters may
have been larger in the past at their time of formation. We color-code those smaller than
1.2RJup in black and see that almost all are beyond a/aRoche > 2. These hot Jupiters would
not have been in danger of tidal disruption even if they were inflated in the past (§3.3).
In the case of in situ formation with no migration at all, we expect the inner semi-major
axis of the hot Jupiter population to occur at the disk edge. The disk edge is thought to
be set by the corotation radius, and therefore hot Jupiters are a factor of several closer to
their stars than expected (e.g., Lee & Chiang 2017). We discuss expectations for the entire
semi-major axis distribution of giant planets formed in situ in §4.1 and §4.3.
As described in §2.2, disk migration may deliver hot Jupiters to half the corotation
period (i.e., the 2:1 resonance with the disk inner edge). These shorter orbital periods
of ∼ 5 days are more consistent with hot Jupiters’ observed orbital periods. See §4.3 for
expectations for the occurrence rate of hot Jupiters (periods < 10 days) relative to warm
(periods 10–200 days) Jupiters from disk migration.
In high eccentricity tidal migration, we expect to see surviving planets at or beyond
2aRoche (Rasio & Ford 1996, Matsumura, Peale & Rasio 2010). When planets begin their
migration at high eccentricities, their initial periapses are approximately half to their final
semi-major axes because afinal = a(1− e2) ≈ 2a(1− e) for e→ 1 (§2.3). Therefore planets
at 2aRoche today must have reached aRoche during high eccentricity migration. Moreover,
if the planets underwent high eccentricity migration before cooling and contracting, their
aRoche during migration would have been larger, so we would see them beyond 2aRoche
today. Although many hot Jupiters are beyond 2aRoche, we also see a population between
1–2 aRoche that high eccentricity migration alone cannot easily account for.
However, in all origins scenarios, subsequent tidal evolution, in which planets raise tides
22
1 100.01
0.10
1.00
a/aRoche
Mp/M
* (10
-3)
Dis
rupt
HEM
dis
rupt
R/RJup: < 1.2> 1.2
1 100.01
0.10
1.00
10.00
P/PRoche
Mp (
Jup)
1 100.50.60.70.91.01.21.51.8
M*
P (days)
Msini > 0.25 MJup
Figure 6 The distribution of hot Jupiters’ orbital periods and semi-major axes tests theories
for their origin. All plotted planets have mass precision of 50% or better. Yellow dotted
line: tidal disruption limit. Red dashed line: tidal disruption limit during high eccentricity
tidal migration (HEM) for e→ 1. Top: Planet-to-star mass ratio vs. semi-major axis scaled
by the Roche radius for planets with 30% precision or better on aRoche. Middle: Same using
planet mass and Roche period (Eqn. 15) for planets with 30% precision or better on PRoche.
Bottom: Stellar mass vs. orbital period of hot Jupiters (including non-transiting planets):
no trend is evident, consistent with halting at the corotation radius. Planets compiled from
exoplanets.org (Wright et al. 2012) and exoplanets.eu (Schneider et al. 2011).
on their stars, can further shrink hot Jupiters’ semi-major axes. Valsecchi, Rasio & Steffen
(2014) find that this subsequent evolution could account for hot Jupiters at 1–2 aRoche for
certain stellar tidal models and parameters. The paucity of very massive hot Jupiters in
www.annualreviews.org • 23
this region (Mp > 3MJup, Fig. 6) may be due to orbital decay (e.g., Damiani & Dıaz 2016).
It is possible that the inner edge hot Jupiters is set not by the origin channel but by aRoche.
See §4.4.3 for a discussion of tidally stripped hot Jupiters than become super-Earths.
One possible way to distinguish among origins scenarios is to look for correlations be-
tween hot Jupiters’ semi-major axes and other parameters. Plavchan & Bilinski (2013)
compared the observed distribution of hot Jupiter semi-major axis vs. stellar mass against
the distribution expected from different origin scenarios. They found that the inner semi-
major axis limit is consistent with scaling as M1/3? , as expected if high eccentricity delivers
hot Jupiters to 2aRoche (Eqn. 15). They ruled out scenarios in which hot Jupiters’ inner
semi-major axis is set by the magnetospheric cavity (i.e., in situ formation at the inner
disk edge or disk migration halted at a 2:1 resonance with the inner disk edge), for which
they expect the inner semi-major axis to scale as M1/7? . However, there are two caveats to
this conclusion. First, as discussed above, the inner edge of the Jupiter distribution may
be sculpted primarily by tidal disruption rather than the formation or delivery location.
Second, if the rotational periods of young stars are largely independent of stellar mass at
the time of gas disk dispersal (Lee & Chiang 2017, references therein), the corotation radius
would follow the same scaling as high eccentricity migration, M1/3? . Panel 3 of Fig. 6 shows
a lack of trend in hot Jupiters’ host star masses vs. hot Jupiters’ orbital periods, consistent
with the inner semi-major axis of the hot Jupiter distribution scaling with M1/3? .
Hot Jupiters’ semi-major axis distribution may have contributions from multiple origins
channels. Recently Nelson, Ford & Rasio (2017) modeled the semi-major axis distribution of
hot Jupiters as resulting from two migration channels, disk migration and high eccentricity
migration. They found that the Kepler and radial velocity sample of hot Jupiters can
be accounted for by high eccentricity migration alone but that hot Jupiters discovered by
WASP and HAT survey – which are limited to shorter orbital periods – need a ∼ 35%
contribution from disk migration.
In summary, the semi-major axes of the closest hot Jupiters appear most consistent with
disk migration, which we expect to deliver to hot Jupiters interior to young stars’ corotation
radii. They appear less consistent with in situ formation, which we expect to deliver hot
Jupiters beyond young stars’ corotation radii, and high eccentricity tidal migration, which
we expect to deliver hot Jupiters to beyond 2aRoche. However, we cannot definitively rule
out in situ formation or high eccentricity tidal migration because tides raised on the star
could move hot Jupiters to shorter orbital periods for certain tidal parameters.
3.5. Ages of hot Jupiter hosts
If hot Jupiters form in situ or arrive via gas disk migration, they should be in place by time
the gas disk dissipates. In contrast, hot Jupiters can arrive through high eccentricity tidal
migration throughout a star’s lifetime. Measuring host star ages may help distinguish among
origin scenarios. However, this approach has not yet made any definitive breakthroughs in
identifying hot Jupiters’ predominant origins channel. Here we summarize the theoretical
and observational challenges to identifying hot Jupiters’ origins using stellar ages.
3.5.1. Expected age distinctions. Although hot Jupiters can complete their high eccen-
tricity migration throughout a star’s lifetime, we expect the bulk of them to arrive early.
Once the mechanisms for generating high eccentricities turn on, they tend to work quickly.
Planet-planet scattering can be triggered by dissipation of the gas disk. If planets remain
24
stable through the dissipation of the gas disk, their subsequent instability timescales are
drawn from a log distribution set by the planets’ spacings (e.g., Chambers, Wetherill & Boss
1996): only particular special spacings result in the system going unstable on a timescale
of order the stellar age. Timescales for eccentricity excitation via secular interactions typ-
ically range from thousands to millions of years depending on the perturber’s mass and
distance. Once the proto-hot Jupiter attains its high eccentricity orbit, the tidal circular-
ization timescale scales as a8final (Eqn. 12). Therefore the possible circularization timescales
span many orders of magnitude and only a very special value for the proto-hot Jupiter’s ini-
tial eccentricity will enable it to tidally migrate on a timescale similar to the star’s lifetime.
Otherwise, if it tidally migrates at all, it will do so quickly. The expectation that most hot
Jupiters will arrive early via high eccentricity migration weakens the distinction between
high eccentricity tidal migration and other origins scenarios. Only with a sample of very
young stars or with a very large sample of main sequence stars can we hope to distinguish
between high eccentricity migration vs. disk migration or in situ formation.
3.5.2. Constraints from young stars. Although the distinction in timescales between the
scenarios is not as dramatic as we might hope, very young stars have the potential to
powerfully distinguish between disk (in situ formation, disk migration) vs. post-disk (high
eccentricity migration) mechanisms. If T Tauri stars, which still have their gas disks, host
hot Jupiters, we can conclude that at least some form in situ or arrive via disk migration.
However, it is challenging to detect and confirm planets orbiting these active young stars.
Recently two hot Jupiters have been discovered orbiting Tauri stars using spectropolarime-
try (Donati et al. 2016, Yu et al. 2017), which are very challenging to explain via high
eccentricity migration. Ongoing surveys will help constrain the occurrence rate of planets
orbiting Tauri stars, allowing us to evaluate whether all hot Jupiters might be delivered
during the gas disk stage.
Next to T Tauri stars, young clusters are the best population to survey for young hot
Jupiters. Hot Jupiters have recently been discovered in young clusters, including two hot
Jupiters with 2 and 4 day orbital periods in the metal-rich 800 Myr Beehive cluster (Quinn
et al. 2012) and HD 285507 b, an 6-day hot Jupiter with e = 0.09 ± 0.02 in the 600 Myr
metal-rich Hyades cluster (Quinn et al. 2014). All three hot Jupiters could be plausibly
explained by any of three origins scenarios; the stars are old enough and the orbital periods
are short enough that high eccentricity tidal migration may have operated. In that case, HD
285507 b may be at the end of its journey, still tidally circularizing. However, its eccentricity
is low enough to have been excited in situ by scattering (Eqn. 10, secular perturbations by
a nearby companion (if a sufficiently massive one is nearby, e.g., Eqn. 14), or possibly even
the disk (Eqn. 8). Most observable open clusters are not young enough to definitively rule
out high eccentricity migration for a given planet.
However, even if we cannot determine the origin of a particular hot Jupiter discovered
in a young cluster, we may be able to distinguish among origins theories by comparing
the occurrence rate of hot Jupiters in young clusters vs. in the field. If high eccentric-
ity tidal migration is at work, the occurrence rate of hot Jupiters will be lower in young
clusters (i.e., because planets with eccentricity excitation timescales and/or tidal circular-
ization timescales of ∼ 1− 10 Gyr will not yet have arrived). Unfortunately, another factor
complicates this expectation: the dynamical environment of the cluster. Encounters with
other stars in the cluster can disturb the planetary system, triggering high eccentricity tidal
migration (e.g., Hao, Kouwenhoven & Spurzem 2013, Shara, Hurley & Mardling 2016; Chat-
www.annualreviews.org • 25
terjee et al. 2012 find that this would not be a common channel in NGC 6791). Brucalassi
et al. (2017) find the occurrence rate for hot Jupiters in the solar-metallicity, solar-age open
cluster M67 is actually higher than in the field (though the discrepancy is only marginally
statistically significant). The addition of hot Jupiters from stellar fly-bys in open clusters
could compensate for missing young hot Jupiters that have yet to undergo migration.
Hot Jupiters have also been discovered orbiting about half a dozen A stars (to date),
which have main sequence lifetimes ∼ 1 Gyr and thus tend to be younger. For example,
hot Jupiter WASP-33b, discovered via Doppler tomography, orbits a star younger than
400 Myr (Collier Cameron et al. 2010). However, comparing the properties of hot Jupiters
orbiting A stars to those orbiting older FGK stars is complicated by the differences in stellar
mass and stellar tidal properties.
3.5.3. Constraints from middle age field stars. Combining Gaia parallax measurements
(Gaia Collaboration et al. 2016) with rotation periods and/or asteroseismology from TESS
(Campante et al. 2016) and PLATO (Rauer et al. 2014) should expand the sample of hot
Jupiters with stellar age estimates. The vast majority of field stars are not young enough
for us to check whether the hot Jupiter arrives while the proto-planetary disk is present.
However, a large sample of field stars with stellar ages would allow us to test for trends.
Two key age trends to investigate are the hot Jupiter occurrence rate and period distri-
bution. From disk migration or in situ formation we expect the occurrence rate and period
distribution to have no dependence on stellar age. In contrast, if high eccentricity tidal
migration is the predominant channel, hot Jupiter occurrence rates should increase with
age and the orbital period distribution should extend to longer orbital periods with age. As
we discussed in §3.5.1, we expect most hot Jupiters formed by high eccentricity migration
to arrive early, so a large sample size is necessary to catch the late comers and identify
differences in their distribution of orbital periods.
Furthermore, high eccentricity tidal migration leads to different eccentricity vs. semi-
major axis distributions over time. Older stars should have circular hot Jupiters out to wider
separations because the hot Jupiters have longer to tidally circularize. Quinn et al. (2014)
compared stellar ages to hot Jupiters’ tidal circularization timescales and found that those
with ages longer than the circularization timescale had significantly larger eccentricities.
However, since the tidal circularization timescale is very sensitive to afinal ( §2.3), the
observed trend between eccentricity and tidal circularization timescale might reflect an
increase in eccentricity with semi-major axis (i.e., still be statistically significant ignoring
the stellar age). A trend in eccentricity vs. semi-major axis may be caused by in situ
eccentricity excitation (§3.1). Comparing eccentricity vs. stellar age within each semi-major
axis interval could help distinguish between tidal circularization vs. in situ excitation.
We caution that robustly identifying trends in a main sequence sample is challenging.
Eccentricity excitation and tidal circularization timescales are drawn from log distributions,
so the main sequence sample (e.g., with a typical ∼ 1 − 10 Gyr age range) is only a small
fraction of the dynamic range of timescales. Moreover, trends with stellar age can be difficult
to distinguish from trends with other stellar properties. For example, Triaud (2011) reported
a trend of host star obliquity decreasing with stellar age, but the trend may instead be with
host star effective temperature (e.g., Winn et al. 2010).
3.5.4. Summary. As of today, stellar ages provide no conclusive evidence regarding the ori-
gins of hot Jupiters. However, investigating how the occurrence rates and properties of
26
hot Jupiters change with stellar age is an interesting area for future study. We recom-
mend investigations of field stars with larger sample sizes of hot Jupiters discovered by
CHEOPS (Broeg et al. 2013), TESS (Ricker et al. 2015), and PLATO (Rauer et al. 2014)
and constraints on stellar properties and ages from TESS (Campante et al. 2016), Gaia
(Gaia Collaboration et al. 2016), and PLATO; continued searches for hot Jupiters orbit-
ing T Tauri stars; efforts to disentangle the effects of age vs. cluster environment on the
occurrence rate of hot Jupiters in open clusters; searches for longer period hot Jupiters in
open clusters; and the development of statistical approaches to better distinguish whether
a trend is due to stellar age or a different stellar property.
3.6. Atmospheric properties of hot Jupiters
The species present in a hot Jupiter’s atmosphere are clues to where and how the hot
Jupiter formed (e.g., Madhusudhan, Amin & Kennedy 2014). The composition of gas and
grains in a proto-planetary disk varies radially as the disk temperature drops and volatiles
condense (e.g., Oberg, Murray-Clay & Bergin 2011). The chemical composition of the gas
and solids the planet accretes changes across snow lines (volatile condensation fronts). For
example, the formation location of a hot Jupiter relative to the water, carbon dioxide, and
carbon monoxide snow lines affect the atmosphere’s C/O ratio. A hot Jupiter formed in
situ would have a composition characteristic of the inner disk – where very few ices can
exist – while a hot Jupiter arriving through high eccentricity tidal migration would have a
composition reflective of the outer disk. A hot Jupiter that underwent disk migration may
have an intermediate composition if it accreted gas along the way (e.g., Alibert et al. 2005).
However, disk dynamics and chemistry can cause ice line locations to vary by an order
of magnitude depending on the disk conditions (e.g., Piso et al. 2015). Uncertainty in
disk conditions makes it challenging to back out a planet’s formation location from its
atmospheric properties. For example, we may deduce from a planet’s low C/O atmosphere
that it formed within the water ice line but suffer from an order of magnitude uncertainty
in where that water ice line was located. Ongoing observations of proto-planetary disks
with ALMA may provide us with the better understanding of realistic disk conditions and
parameters necessary to pin down typical snow line locations. More pessimistically, ALMA
may reveal too much diversity in disk conditions to ever infer a typical snow line.
Another uncertainty is the extent to which planetesimals get mixed into giant planets’
atmospheres. This uncertainty complicated the interpretation of Sing et al. (2016)’s hot
Jupiter characterization survey. Sing et al. (2016) found that hot Jupiter atmospheric
compositions are consistent with no primordial water depletion relative to nebular gas
containing water in vapor form. One possibility is that hot Jupiters formed within the
water ice line. A second possibility is that they formed beyond the ice line and underwent
disk migration but accreted gas along the way. A final possibility is that they formed
entirely beyond the snow line and arrived via migration but icy planetesimals were oblated
in their atmospheres, replenishing the water fraction.
Another challenge to using atmospheric properties to test hot Jupiter origin hypotheses
is the difficulty in measuring species in the atmospheres. The C/O ratio is one example of
an atmospheric quantity strongly influenced by location relative to ice lines yet challenging
to measure. It can be inferred from the abundance of carbon monoxide (depleted at high
C/O ratio) and methane (enhanced at high C/O ratio). However, C/O inferences have
proven to be sensitive to which datasets and analysis techniques were used (e.g., Kreidberg
www.annualreviews.org • 27
et al. 2015) and to which priors were imposed dictating chemically-possible atmospheres
(e.g., Heng & Lyons 2016). Higher signal-to-noise spectra observed by the upcoming James
Webb Space Telescope may resolve some of these discrepancies by shifting the analysis into
a more data-driven regime.
4. TESTING THE ORIGIN THEORIES BY CONNECTING HOT JUPITERSTO OTHER POPULATIONS
Hot Jupiters’ origins manifest not only in their intrinsic properties (§3) but also in their
connections to other exoplanet populations. Here we review how the following connections
square with hot Jupiters’ origins: the occurrence rate of hot Jupiters vs. more distant giant
planets (§4.1), companions of hot Jupiters (§4.2), properties of hot vs. warm Jupiters (§4.3,
and properties of hot Jupiters vs. smaller planets (§4.4).
4.1. Hot Jupiter occurrence rates relative to wider separation giant planets
4.1.1. Overall relative occurrence rates. About one in ten giant planet systems contains a
hot Jupiter (Howard et al. 2010, Mayor et al. 2011, cf Wright et al. 2012 and Guo et al.
2017 regarding the apparent discrepancy between RV surveys and Kepler HJ occurrence
rates), with the giant planet occurrence rate dropping sharply within 200 days (Fig. 4).
The occurrence rate of hot Jupiters relative to wider separation giant planets reflects the
efficiency of their origins channel. If they formed in situ (§2.1), their occurrence rate is set
by the propensity of disks to form giant planets close to their stars vs. at wider separations.
If they formed ex situ, their occurrence rate is set by the efficiency of transporting giant
planets close to their star by disk migration (§2.2) or high eccentricity tidal migration (§2.3).
The in situ formation hypothesis (§2.1) currently lacks a clear prediction for hot Jupiters’
relative occurrence rate. The ease of forming hot Jupiter depends on the inner disk’s local
solid surface density and, to a lesser extent, the gas opacity (e.g., Lee & Chiang 2016).
Building a core capable of runaway gas accretion requires a sufficient amount solids close to
the star. Equation 6 dictates a factor 10,000 higher surface density to build a massive core
at a 3 day orbital period than at a 3000 day orbital period; whether this factor is achievable
depends on the disk’s mass in solids and their radial distribution. The radial distribution
in turn depends on the (still not well quantified) efficiency of radial transport of pebbles,
planetesimals, and embryos and the extent to which solids can pile up in one location. As we
will discuss in §4.4, a complementary avenue to evaluating the in situ formation hypothesis
is to compare giant planet occurrence rate vs. orbital period to super-Earth occurrence
rates vs. orbital period.
Disk migration (§2.2) can deliver hot Jupiters at the observed relative rate for plausi-
ble migration parameters. For example, Coleman & Nelson (2016) found good agreement
between their simulations of giant planet migration and the observed occurrence rates.
Jupiters’ final locations depend on the migration timescale, which may span many orders of
magnitude depending on disk conditions, and the remaining disk lifetime after the Jupiter’s
formation. If the migration timescale is much longer than the disk lifetime, the Jupiter
will not stray far from its birthplace. If the remaining disk lifetime is much longer than
the migration timescale, the Jupiter can become a hot Jupiter (assuming some mechanism
halts its migration prior to tidal disruption, §2.2).
In the high eccentricity tidal migration scenario (§2.3), hot Jupiters’ relative occur-
28
Table 2. Theoretical efficiencies of forming hot Jupiters from cold Jupiters via high
eccentricity tidal migration. The observed ratio of hot to cold Jupiters is ∼ 1 : 10.
Mechanism Study HJ/J (%) Assumptions of study
Stellar binary Kozai a ∼1–3 System of single giant planet and bi-nary companion
Planet secular coplanar b ∼3–5 System of two giant planet; HJ beginsat 1 AU
Planet-planet Kozai c ∼ 5 System of two giant planets; mutualinclinations drawn from an isotropicdistribution; HJ begins at 1 AU
Planet-planet scattering d ∼3–5 System of at least three giant planets;system goes unstable.
Secular chaos e ? Not quantified due to strong depen-dence on uncertain initial conditions
Note. — a: Munoz, Lai & Liu (2016), b: Petrovich (2015a), c: Petrovich & Tremaine(2016), d: Beauge & Nesvorny (2012), e: Wu & Lithwick (2011).
rence rate depends on the efficiency of the mechanism that raises the proto-hot Jupiter’s
eccentricity. Forming a hot Jupiter requires driving the eccentricity high enough for tidal
dissipation to be effective but not so high that the proto-hot Jupiter hits the star or is tidally
disrupted. Investigations of high eccentricity tidal migration have found low rates of hot
Jupiter production inconsistent with observations, even with optimistic assumptions, as we
summarize in Table 2. A few outstanding theoretical issues remain in understanding in the
hot Jupiter efficiency from high eccentricity tidal migration. First, the secular excitation
mechanisms (§2.3) – stellar binary Kozai, planet-planet secular coplanar, and planet-planet
Kozai – can easily be shut off by bodies between the proto-hot Jupiter and its perturber
(e.g., an additional giant planet, a nearby small planet). Accounting for such planets would
reduce the efficiency further. Second, more investigation of how the results depend on initial
conditions is necessary. For example, beginning a proto-hot Jupiter at 1 AU (e.g., Petrovich
& Tremaine 2016) requires less eccentricity excitation than starting it further from the star.
Third, more investigation is needed on whether hot Jupiters can be saved from tidal disrup-
tion, a major sink of hot Jupiters in high eccentricity migration models. For example, Wu
(2017) recently showed that giant planets can undergo rapid tidal circularization through
f-mode dissipation. This rapid circularization may allow a hot Jupiter to safely migrate and
decouple from its perturber before its eccentricity is raised high enough for tidal disruption.
Finally, the interplay among all the different planet-planet excitation mechanisms should be
explored further (e.g., Nagasawa & Ida 2011); it is unclear whether hot Jupiter formation
rates from different planet-planet eccentricity excitation mechanisms should be summed.
4.1.2. Relative occurrence rate vs. planet mass. Hot Jupiters have lower masses than their
more distant counterparts (Fig. 6); in particular, we observe a dearth of giant planets above
>∼ 3MJup at orbital separations between the Roche limit and twice the Roche limit. One
www.annualreviews.org • 29
possibility is that lack of massive hot Jupiters is not a result of hot Jupiters’ origin channel
but of hot Jupiters raising tides on their host stars. More massive planets can more quickly
transfer their angular momentum to spin up their stars and undergo tidal decay, ultimately
leading to tidal disruption.
Alternatively, hot Jupiters’ lower masses may be primarily established by their ori-
gins channel. The in situ formation hypothesis may preferentially produce lower mass hot
Jupiters (e.g., Batygin, Bodenheimer & Laughlin 2016). A giant planet’s core can either
form as an isolation mass from a narrow feeding zone or grow via collisions of smaller cores
from a wider annulus as the gas surface density declines (e.g., Boley, Granados Contreras
& Gladman 2016). The latter scenario can operate with a lower local solid surface density
(helping mitigate an obstacle to in situ formation) but can stunt the hot Jupiter’s growth
by giving it less time to accrete gas and less gas to accrete (e.g., Lee & Chiang 2016).
Migration is also consistent with lower masses for hot Jupiters. Less massive hot Jupiters
may migrate more efficiently because they are less effective at opening deep gaps in the
disk that would slow their migration. Masset & Papaloizou (2003) found that “runaway
migration,” in which a feedback caused by corotation torques leads to rapid inward mi-
gration,operates most effectively for planets just below Jupiter mass. Sub-Jovians are sus-
pectible to runaway migration because they are less massive than the surrounding disk yet
massive enough to open shallow gaps in the disk that facilitate the corotation torque.
It is currently unclear whether lower masses for hot Jupiters are consistent with high
eccentricity tidal migration. Lower mass planets are more easily disturbed onto elliptical
orbits by other planets in planet-planet mechanisms (Naoz et al. 2011, Wu & Lithwick 2011,
Petrovich 2015a). However, they are also more easily tidally disrupted (Anderson, Storch
& Lai 2016, Munoz, Lai & Liu 2016).
4.1.3. Relative occurrence rate vs. stellar properties. The overall giant planet occurrence
correlates with stellar mass (Johnson et al. 2010; peaking at 2 M�, Reffert et al. 2015)
and metallicity (Gonzalez 1997, Santos, Israelian & Mayor 2001, Santos et al. 2003, Santos,
Israelian & Mayor 2004, Fischer & Valenti 2005, Sousa et al. 2011, Guo et al. 2017). These
correlations are thought to reflect an underlying connection between giant planet occurrence
rate and disk mass, with more massive, solid-rich disks spawning cores massive and quick-
growing enough to accrete gas during the gas disk lifetime (§2.1). The disk mass has been
observationally linked to the host star mass (Ansdell et al. 2016, 2017, Pascucci et al. 2016);
it is still unclear whether the disk mass in solids is correlated with host star metallicity
(Moro-Martın et al. 2015, Gaspar, Rieke & Ballering 2016).
Different origins channels have different predictions for whether the occurrence rate of
hot Jupiters relative to other giant planets should depend on stellar mass and metallicity.
The efficiency of forming hot Jupiters in situ relative to more distant giant planets is strongly
dependent on the local solid surface density (§2.1), which is controlled by both the total disk
mass in solids and the solid transport efficiency. The former should produce a correlation
between hot Jupiters and host star metallicity/mass. A higher solid surface density would
allow planets to form more quickly in the outer disk. A faster formation time would give a
Jupiter more time to migrate, increasing the ratio of hot to cold Jupiters. However, given
the many orders of magnitude of possible migration speeds (i.e., migration speeds can be
much longer or much shorter than the disk lifetime), the increase may not be significant.
In the high eccentricity tidal migration channel, the relative hot Jupiter occurrence
rate vs. stellar mass/metallicity depends on the mechanism for raising the hot Jupiter’s
30
eccentricity. In the case of high eccentricity migration triggered by a stellar perturber, we
do not expect a correlation with host star metallicity/mass. We may even expect an anti
correlation because if a disk spawns too many giant planets, they may protect each other
from the secular effects of the stellar companion. In contrast, high eccentricity migration
triggered by a planetary perturber is likely to produce a correlation. A more massive and/or
solid enriched disk can spawn giant planets at more locations (Fig. 2), allowing multiple
giant planets to form.
Observationally, there is some evidence for a further enhancement of hot Jupiters relative
to cold Jupiters at higher host star metallicities. Jenkins et al. (2017) found that giant
planets with orbital periods less than 100 days orbit more metal rich stars than giant
planets with orbital periods greater than 100 days. Dawson & Murray-Clay (2013) found a
larger difference in occurrence rates for hot Jupiters orbiting metal rich vs. metal poor stars
than for longer period planets orbiting metal rich vs. metal poor stars. The correlation
with stellar mass is less certain. Obermeier et al. (2016) found the occurrence rate of hot
Jupiters orbiting M-stars vs. FGK stars could be the same, but for the former, they only
have an upper limit.
4.2. Companions of hot Jupiters
The hot Jupiters origin theories make different predictions for whether hot Jupiters are
likely – or required – to be members of multi-planetary or binary systems and the mass and
proximity of these companions. In situ formation tends to spawn hot Jupiters accompanied
by nearby planets (e.g., Hansen & Murray 2013, Boley, Granados Contreras & Gladman
2016), which are not necessarily in or near orbital resonance. Disk migration tends to
deliver hot Jupiters with resonant companions (Malhotra 1993, Lee & Peale 2002, Raymond,
Mandell & Sigurdsson 2006), which may be giant or small. Pairs of small planets can
plausibly escape from resonance (Goldreich & Schlichting 2014), but orbital resonances
involving one or more giant planets tend to persist. Additionally, small, non-resonant
planets might be able to form in situ nearby after the gas giant’s migration. In contrast,
tidal migration wipes out small planets inside the Jupiter’s initial orbit (e.g., Mustill, Davies
& Johansen 2015) but requires a stellar or giant planet companion to trigger the high
eccentricity. If planet-planet scattering produces the high eccentricity, the companion could
be ejected but for secular mechanisms, the companion should still be present and have
the necessary properties to have triggered high eccentricity migration. In summary, we
expect nearby planets in the in situ formation scenario, nearby resonant planets in the disk
migration scenario, and distant companions in the high eccentricity migration scenario.
4.2.1. Distant companions of hot Jupiters. Several follow-up studies of hot Jupiters have
been conducted to search for both stellar and planetary companions (Fig. 7). The Friends
of Hot Jupiters survey probed companions using long baselines of radial velocities (Knutson
et al. 2014, Bryan et al. 2016) sensitive to giant planet companions out to 10s of AU, infrared
spectroscopy (Piskorz et al. 2015), and direct imaging (Ngo et al. 2016, 2017) sensitive to
low-mass stars at 10s to 1000s of AU. Other surveys for distant companions include Endl
et al. (2014), Evans et al. (2016), and Neveu-VanMalle et al. (2016).
The properties of hot Jupiters’ stellar companions are incompatible with high eccen-
tricity migration triggered by stellar Kozai being the dominant channel of formation. Ngo
et al. (2016) found an upper limit of 16±5% on hot Jupiters with stellar companions capa-
www.annualreviews.org • 31
1 10 100 1000a (AU)
0.1
1.0
10.0
100.0
1000.0
frien
d m
ass
(Jup
iters
)Ove
rcome G
R, P HJ=10
day
Kozai
excite
e, P Ju
p,0=4 y
r
Stars (AO), Ngo+16Planets (RV)RV Trend
Figure 7 Mass vs. semi-major axis of detected companions to hot Jupiters. Diamonds
indicate hot Jupiter companions to detected via the radial velocity method (red; compiled
from Knutson et al. 2014, Wright et al. 2012) or direct imaging (blue; Ngo et al. 2016).
Dotted lines represent radial velocity trends. Companions to the top right the red dashed
line are capable of overcoming GR precession to excite a hot Jupiter’s eccentricity in situ
at a 10 day orbital period (Eqn. 14); those to the top right of the solid like are capable of
exciting the eccentricity a proto-hot Jupiter on a 4 year orbital period high enough for the
proto-hot Jupiter to circularize to a 5 day orbital period (Eqn. 13). Most hot Jupiters do
not have a companion capable of overcoming GR to raise their eccentricity through Kozai-
Lidov at their present day short orbital periods (i.e., detection or upper limit is below red
dashed line) or of triggering high eccentricity migration of a proto-hot Jupiter from beyond
the ice line (i.e., points below the solid red line). See §2.3.3 for a discussion of other relevant
timescales, such as tidal precession.
ble of triggering high eccentricity tidal migration via Kozai–Lidov cycles (if the hot Jupiter
formed between ∼1–5 AU). Most companions are not massive and nearby enough to over-
come general relavistic (Eqn. 14) and tidal precession. Ngo et al. (2016) find no correlation
between binary companions and spin-orbit misalignments.
In contrast, hot Jupiters’ planet companions are generally nearby and massive enough to
secularly trigger high eccentricity migration if they have the necessary eccentricity and/or
mutual inclination. Knutson et al. (2014), Bryan et al. (2016) found that 70±8% of hot
Jupiters have an outer planet between 1–20 AU and 1–13 Jupiter masses. These outer
planets are capable of having raised the eccentricities of proto-hot Jupiters formed ex situ
between ∼1–5 AU, leading to high eccentricity tidal migration. How do these companions
square with the population of eccentric hot Jupiters (§3.1)? They are consistent with the
theory that eccentric hot Jupiters are those finishing their high eccentricity migration. Most
are not nearby and massive enough (Fig. 7) to having excited the hot Jupiter’s eccentricity
in situ (i.e., if the hot Jupiter formed in situ or underwent disk migration). No eccentric
hot Jupiter has a known companion – or set of possible companions based on a linear
radial-velocity trend – capable of exciting the hot Jupiter’s eccentricity in situ. However,
for some eccentric hot Jupiters, the limits on companions are weak and more long baseline
observations are needed to rule out the presence of such a companion.
32
0.1 1 100.01 a (AU)
GJ-876
Kepler-9
Kepler-30
55 Cnc~3:1
WASP-47 ~2:1
2:1 2:1
~2:1
~2:1 ~7:3
Kepler-89~2:1 ~2:1
Figure 8 Warm and Hot Jupiters for which the companions also have semi-major axes less
than 1 AU. Orbital resonances are labeled and sizes scale approximately with log planet
mass. Hot Jupiter WASP-47b (Becker et al. 2015) may be in the short period tail of a class
of system featuring a close-in giant planet in orbital resonance with one or more neighbors.
This architecture often also includes an ultra-short period super-Earth.
4.2.2. Nearby companions of hot Jupiters. Nearby planets are generally absent in systems
containing hot Jupiters. Latham et al. (2011) and Huang, Wu & Triaud (2016) found that
hot Jupiters are much less likely to have other transiting planets in their systems. Steffen
et al. (2012) used the lack of transit timing variations of hot Jupiters to rule out nearby
planets, even at low masses. However, there are exceptions to the trend that hot Jupiters
lack nearby planets. Millholland, Wang & Laughlin (2016) found phase curve evidence for a
non-transiting hot Jupiter in a system containing a 150 day orbital period candidate mini-
Neptune. If confirmed, this system would be incompatible with high eccentricity migration
of the hot Jupiter from beyond 1 AU (e.g., Mustill, Davies & Johansen 2015). The most
striking exception is WASP-47b (Fig. 8), a hot Jupiter near a 2:1 resonance with an outer
Neptune (Becker et al. 2015). The WASP-47 system also has an ultra-short period super-
Earth. WASP-47b bears resemblance to a number of systems featuring warm Jupiters with
resonant neighbors, which are depicted in Fig. 8 and discussed further in §4.3.
Although hot Jupiters’ lack of nearby planets has generally been interpreted as evidence
for high eccentricity migration, two sets of recent studies have argued that disk migration
and in situ formation could be compatible with this observed trend. In the first set of studies,
www.annualreviews.org • 33
Ogihara, Inutsuka & Kobayashi (2013), Ogihara, Kobayashi & Inutsuka (2014) suggested
that super-Earths cannot form near a hot Jupiter that migrated or formed in situ. Super-
Earths would tend to drive the hot Jupiter into its star, provided that disk conditions did
not enable hot Jupiters to open wide gaps. This theory does not predict a complete absence
of nearby planets: low mass terrestrial planets could co-exist with the hot Jupiter. A more
quantitative comparison of the surviving systems to the Steffen et al. (2012) limits from
transit timing variations would clarify whether low mass terrestrial planets would necessarily
escape detection. Furthermore, in this scenario the hot Jupiters we observe today would be
the exceptional survivors in systems where other planets never formed near the hot Jupiter
or accompanied them on their migration. More work is necessary to explore whether a
sizable population of hot Jupiters could form and remain lonely.
In the second study, Schlaufman & Winn (2016) argued that the population statistics
of hot Jupiter companions belay high eccentricity tidal migration from beyond the ice
line. They found that the occurrence rate of longer period giant planet companions to
hot Jupiters is consistent with that of companions to longer period Jupiters (periods > 10
days). Their results hinge on two hot Jupiters with sub-AU companions discovered via radial
velocity: upsilon Andromeda b (at 0.83 AU) and HIP 14810 (at 0.5454 AU). Companions
in the Schlaufman & Winn (2016) sample are at ∼ 0.5–3 AU and are not incompatible with
other studies investigating hot Jupiter companions at much closer (Latham et al. 2011,
Steffen et al. 2012, Huang, Wu & Triaud 2016) or wider (e.g., Friends of Hot Jupiters)
separations. Steffen et al. (2012), Latham et al. (2011), and Huang, Wu & Triaud (2016)
are sensitive to small, nearby planets that would not have been detectable by radial velocity.
Conversely, the Schlaufman & Winn (2016) systems feature longer period planets that would
be unlikely to transit and potentially too widely separated to cause TTVs. As part of the
Friends of Hot Jupiters survey, Bryan et al. (2016) found that the occurrence rate for hot
Jupiter companions is larger than for warm Jupiter companions, but these companions are
typically at wider separations than Schlaufman & Winn (2016)’s sample.
Schlaufman & Winn (2016)’s finding – that hot Jupiters’ companion rate is consistent
with that of the overall population – is not necessarily in tension with high eccentricity tidal
migration. The efficiency of producing hot Jupiters through high eccentricity tidal migration
is low and hot Jupiters are likely special outcomes of dynamical processes that affect many
systems (§4.1). Therefore we might expect the outer architectures of hot Jupiter systems to
resemble those of giant planet systems without hot Jupiters. However, the question remains
whether having a hot and warm Jupiter together in the same system is incompatible with
high eccentricity tidal migration. Such an architecture would indeed seem to require that
the hot Jupiter formed or migrated to the warm Jupiter region before the hot Jupiter high
eccentricity tidal migration commenced; otherwise the warm Jupiter would be disturbed
during hot Jupiter’s tidal migration.
4.2.3. Summary of hot Jupiters’ companions. Overall, the companions of hot Jupiters give
most support to the origin hypothesis of high eccentricity tidal migration triggered by a
planet companion. The majority of hot Jupiters are accompanied by long period planetary
companions that are massive and nearby enough to have secularly triggered the hot Jupiter’s
high eccentricity tidal migration. (However, we do not yet know whether the companions
have necessary eccentricities and inclinations. Also, such companions are not incompatible
with disk migration or in situ formation.) In contrast, the majority of hot Jupiters are not
accompanied by stellar companions capable of triggering their high eccentricity migration,
34
implying that stellar Kozai triggered by high eccentricity migration is not a predominant
channel. Most hot Jupiters are not accompanied by nearby planets, which we would expect
to be present under the in situ formation and disk migration hypotheses and absent under
the high eccentricity migration hypothesis. However, WASP-47b (Becker et al. 2015) cannot
be explained by high eccentricity tidal migration and would need to arrive via a different
channel.
4.3. Hot and Warm Jupiters
Warm Jupiters are giant planets orbiting close to their star but at wider separations than
hot Jupiters. Their longer orbital periods, ∼10-200 days, made them challenging to discover
in ground-based transit surveys (see Gaudi, Seager & Mallen-Ornelas 2005 for a discussion
of ground-based surveys’ selection effects), and hence they have received less attention
than hot Jupiters. Like hot Jupiters, warm Jupiters could have originated from in situ
formation, disk migration, or high eccentricity migration. Also like hot Jupiters, their semi-
major axes are too small for them to have been scattered directly from several AU (a factor
of 10 change in energy; e.g., Dong, Katz & Socrates 2014). Fig. 4 shows how different
mechanisms populate the warm Jupiter region.
However, theoretical studies of the formation and orbital evolution of hot Jupiters have
found it challenging to account for the occurrence rates, eccentricities, and other properties
of warm Jupiters. This challenge may indicate a major problem in our understanding of
hot Jupiters or that hot and warm Jupiters do not have a common origin. Distinguishing
between these two possibilities is key to a complete understanding of hot Jupiters.
4.3.1. Hot vs. Warm Jupiter occurrence rates. The occurrence rate per log interval of
giant planets dips in the warm Jupiter region by a factor of a few (Fig 4). This feature is
known as the Period Valley (Jones et al. 2003, Udry, Mayor & Santos 2003, Wittenmyer
et al. 2010, Santerne et al. 2016). See Wittenmyer et al. 2010 for evidence of the statistical
significance of the Period Valley, whose width, depth, and dependence on stellar properties
are still being investigated (e.g., Santerne et al. 2016). Although a Period Valley could be
compatible with in situ formation if solids piled up in the innermost disk, we will discuss
in §4.4 that this explanation seems incompatible with super-Earth occurrence rates vs.
orbital period. The Period Valley feature is plausibly compatible with disk migration. If a
hot Jupiter’s migration timescale is comparable to the disk lifetime after the hot Jupiter’s
formation, the Jupiter can get stranded mid-migration as a warm Jupiter. Therefore warm
Jupiters are less common than hot or cold Jupiters because they require a special migration
timescale (i.e., particular disk conditions). Another possibility is that photo-evaporation
of the inner disk leaves later-forming giant planet stranded beyond ∼ 100 days (Alexander
& Pascucci 2012), unable to migrate. Coleman & Nelson (2016) reproduced the period
distribution of giant planets using a particular set of disk parameters.
Although the occurrence rate of warm Jupiters per log interval is lower than for hot
Jupiters, the total number of warm Jupiters is larger. Studies of high eccentricity tidal
migration have severely under-produced warm Jupiters relative to hot Jupiters: in other
words, they create a Period Valley that is far too deep. In the high eccentricity migra-
tion origins theory, warm Jupiters are proto-hot Jupiters in the midst of migration. Most
high eccentricity migration mechanisms severely underproduce warm Jupiters (e.g., Wu &
Lithwick 2011, Beauge & Nesvorny 2012, Petrovich, Tremaine & Rafikov 2014, Petrovich
www.annualreviews.org • 35
2015a). One way to enhance the number of warm Jupiters is to invoke perturbers massive
and nearby enough to remain coupled to the proto-hot Jupiter (e.g., Dong, Katz & Socrates
2014). In this scenario, the proto-hot Jupiter’s eccentricity can oscillate so that it spends
less time at periapses small enough for effective tidal circularization, slowing down the mi-
gration. However, Petrovich & Tremaine (2016) found that even this configuration leads to
a ratio of warm Jupiters to hot Jupiters that is a factor of ∼ 5 lower than observed. Dawson
& Chiang (2014) found a dynamical configuration that can further extend the fraction of
time spent as a warm Jupiter but did not quantify the likelihood of this configuration or
its effect on the observed ratio of warm to hot Jupiters. The warm to hot Jupiter ratio is
currently a major weakness in the high eccentricity tidal migration hypothesis, at least if it
were the only channel.
4.3.2. Warm Jupiters: too eccentric for comfort. Warm Jupiters have a wide range ec-
centricities. Their eccentricity distribution contains a low eccentricity component and a
component with an approximately uniform distribution (e.g., Petrovich & Tremaine 2016).
The circular component, which we will discuss in §4.3.3, is a challenge for the high eccentric-
ity tidal migration hypothesis. The eccentric component is challenging for in situ formation
or disk migration. Like for hot Jupiters (§3.1), in situ formation or disk migration leads
to warm Jupiters on low eccentricity orbits that cannot be excited by subsequent scatter-
ing (Petrovich, Tremaine & Rafikov 2014). Anderson & Lai (2017) advocate that warm
Jupiters’ eccentricities could be excited by an outer companion, which is feasible for warm
Jupiters (e.g., Dong, Katz & Socrates 2014). However, Anderson & Lai (2017) find that
this mechanism most easily explains moderately eccentric warm Jupiters, rather than high
eccentricity warm Jupiters, because the fraction of time spent at high eccentricities for an
initially circular warm Jupiter is low. Mustill, Davies & Johansen (2017) identified three
additional mechanisms – each of which involve several steps – to generate warm Jupiters in
systems of three or more planet. However, future work can quantify the efficiency of these
more complex mechanisms.
4.3.3. Warm Jupiters: too circular for comfort. Conversely, the high eccentricity tidal
migration hypothesis has been unable to account for the low eccentricity component of the
warm Jupiter population. Recall that under the high eccentricity migration hypothesis, all
warm Jupiters are in the midst of tidal migration. If the perturber is decoupled from the
migrating proto-hot Jupiter, its periapse should be close enough to the star for tides to
operate (red region of Fig. 4). With a couple exceptions, such as HD 80806b (e = 0.93,
Naef et al. 2001, Wu & Murray 2003), elliptical warm Jupiters have eccentricities too high
for in situ formation or disk migration (§4.3.2) but too low to be undergoing tidal migration
(i.e., as denoted by the white region in Fig. 4).
Dong, Katz & Socrates (2014), Dawson & Chiang (2014), and Petrovich & Tremaine
(2016) modeled warm Jupiters as undergoing tidal circularization but still coupled to a com-
panion on a mutually inclined orbit. We observe warm Jupiters currently at eccentricities
currently too low for tidal circularization but they periodically reach higher eccentricities.
Petrovich & Tremaine (2016) successfully reproduced the uniform eccentricity distribution
of the eccentric component of warm Jupiters using planet-planet Kozai-Lidov driven high
eccentricity tidal migration.
However, in most cases we do not have the constraints on the three-dimensional architec-
tures of warm Jupiter systems necessary to test whether they have the requisite mutually
36
inclined companions. Recently mutual inclinations have been measured for a handful of
warm Jupiter systems using transit time and duration variations, and results have been
mixed. Masuda (2017) found that the three-dimensional architecture of the Kepler-693 sys-
tem is capable of driving high eccentricity migration. Dawson et al. (2014) found that the
massive nearby planet companion to eccentric (e=0.81; Dawson et al. 2012) warm Jupiter
Kepler-419b is co-planar and not capable of driving high eccentricity tidal migration. Daw-
son & Chiang (2014) argued that half a dozen systems of warm Jupiters with nearby massive
companions had the requisite mutual inclinations based on indirect evidence from clustering
in the separations of their argument of periapse. There is currently debate over whether
high eccentricity migration could have taken place in such systems. Antonini, Hamers &
Lithwick (2016), Masuda (2017) argue that rewinding the warm Jupiter’s tidal migration
to an initial semi-major axis > 1 AU would result in a spacing too close to its compan-
ion for stability. In principle, such an instability may be what actually triggered the high
eccentricity migration, but Antonini, Hamers & Lithwick (2016) found in simulations that
high eccentricity migration was an uncommon outcome of scattering in these systems.
Even if we can account for observed eccentricities too low for tidal circularization, a
remaining problem is that the expected population of super-eccentric migrating proto-hot
Jupiters is missing. Hot Jupiters are continuously being spawned throughout the galaxy as
new stars are born, gas giants form, have their eccentricities excited, and tidally migrate.
Socrates et al. (2012) proposed that regardless of the mechanism of eccentricity excitation,
we can test the origins theory of high eccentricity migration by searching for proto-hot
Jupiters on highly elliptical orbits. If moderately eccentric hot Jupiters (0.2 < e < 0.6)
(§3.1, Fig. 4) are those finishing their tidal circularization, we expect the corresponding
super-eccentric proto-hot Jupiters (i.e., those with e > 0.9 and the same afinal, ranging
from 0.04–0.10 AU) dictated by the relative tidal evolution timescale. Socrates et al. (2012)
identified the Kepler sample (see Lissauer, Dawson & Tremaine 2014 for a review of the
Kepler Mission) as particularly well-suited for this search, because the transit probability
of short and long period planets is the same for a given afinal. Unlike most ground-based
transit surveys, Kepler has the baseline and sensitivity to detect warm Jupiters. However,
using an approach to identify super-eccentric Jupiters based on their transit shape and
duration (e.g., Dawson & Johnson 2012), Dawson, Murray-Clay & Johnson (2015) found
a paucity of super-eccentric proto-hot Jupiters in the Kepler sample, inconsistent with the
expectation from high eccentricity tidal migration. Wu (2017) recently proposed that f-
mode tidal dissipation could cause hot Jupiters to migrate very quickly through the highly
eccentric stage, causing us to miss super-eccentric Jupiters. However, this tidal dissipation
mechanism requires afinal . 0.04 AU and therefore can likely only account for a fraction of
the missing super-eccentric Jupiters.
4.3.4. Stellar obliquities of warm Jupiters’ host stars. If warm Jupiters have the same origin
as hot Jupiters, the distribution of stellar obliquities of warm Jupiters’ host stars could
serve as a primordial distribution, unsculpted by tides raised on the star. Fig. 5 depicts
the processes affecting hot Jupiters’ spin-orbit alignments. For warm Jupiters, we can
escape the final confusing step in which the star can get realigned, erasing the distribution
sculpted by hot Jupiters’ origins mechanism. (Recall that tidally migrating warm Jupiters
are experiencing tides raised on the planet. Tides raised on the star, sensitive to a rather
than a(1− e2), likely only become important when hot Jupiter-hood is achieved.)
However, the measurement and interpretation of warm Jupiters’ stellar obliquities is an
www.annualreviews.org • 37
outstanding problem requiring a larger observational sample size. A few preliminary results
have been ambiguous to interpret. Li & Winn (2016) find that planets are more misaligned
at longer orbital periods, consistent with the idea that the primordial distribution has been
sculpted by tides. However, the trend of increasing misalignment vs. orbital period persists
out to 50 days, a much larger distance than we expect tides to operate. Kepler-56, which
hosts two transiting giant planets, is a misaligned cool star (Huber et al. 2013). However,
the misalignment may be caused by a companion torquing the inner pair (e.g., (Li et al.
2014c, Otor et al. 2016)), rather than a primordial disk misalignment.
4.3.5. Companions to warm Jupiters. Warm Jupiters have different companions than hot
Jupiters, which may suggest they have a different origins channel. Huang, Wu & Triaud
(2016) discovered a population of warm Jupiters with nearby super-Earths, which are in-
compatible with high eccentricity tidal migration (Mustill, Davies & Johansen 2015), in the
Kepler sample. These systems exhibit a strong disparity from hot Jupiters, which generally
lack nearby companions (Huang, Wu & Triaud 2016; see also §4.2). Exceptional hot Jupiter
systems like WASP-47 containing nearby planets (§4.2, Becker et al. 2015) may be tail end
members of this warm Jupiter population. Regarding longer period companions, Bryan
et al. (2016) found the companion rate to be lower for warm Jupiters than for hot Jupiters
(49± 10% vs. 75± 5%). If these companions are necessary for high eccentricity migration,
this finding may imply that fewer warm Jupiters came through this origins channel than
hot Jupiters.
4.3.6. Warm Jupiters: multiple origins channels?. None of the three origins channels alone
can account for all of warm Jupiters’ observed properties. One promising option is to invoke
two origins channels, one for circular warm Jupiters and one for eccentric warm Jupiters.
Distinctions in other properties between eccentric vs. circular warm Jupiters support this
hypothesis. Low eccentricity warm Jupiters are less likely to have giant planet compan-
ions (Dong, Katz & Socrates 2014, Bryan et al. 2016) and warm Jupiters that have very
nearby companions incompatible with tidal migration have lower eccentricities (Dawson,
Murray-Clay & Johnson 2015). Warm Jupiters orbiting metal-poor stars are confined to
low eccentricities while those orbiting metal-rich stars exhibit a range of eccentricities (Fig.
4, Dawson & Murray-Clay 2013; as discussed in §3.1). This trend may link the origins of
eccentric warm Jupiters to planet-planet gravitational interactions. Moreover, Dawson &
Murray-Clay (2013) found that moderately eccentric hot Jupiters orbit higher metallicity
stars. If high eccentricity tidal migration is correlated with host star metallicity, the Kepler
sample’s lower metallicity may translate to fewer hot Jupiters originating from high eccen-
tricity migration. A lack of high eccentricity migration could contribute to both the overall
lower occurrence rate of hot Jupiters in the Kepler sample (§4.1.3; but not entirely account
for, see Guo et al. 2017) and the lack of super-eccentric Jupiters (§4.3.3).
4.4. Comparing and connecting hot Jupiters to smaller planets
In the early 2000s, radial-velocity techniques reached the precision necessary to discover
small planets (e.g., Santos et al. 2004). In the 2010s, radial-velocity surveys amassed large
enoughs sample to statistically characterize the properties of small planets (e.g., Howard
et al. 2010, Mayor et al. 2011) and the Kepler Mission discovered an abundance of small
transiting planets (see Lissauer, Dawson & Tremaine 2014 for a review of Kepler results).
38
Comparing hot Jupiters to smaller planets is another possible avenue to test theories for the
origins of hot Jupiters. All three origins channels can produce hot planets of various masses
and sizes, so whatever channel operates likely produces at least some hot super-Earths and
Neptunes. However, the predominant origin channel for hot super-Earths and Neptunes
may be different than for hot Jupiters. Here we review how different hot Jupiter origins
channel contribute smaller planets and if and how we can compare the properties of hot
Jupiters vs. smaller planets to test theories for hot Jupiters’ origins.
4.4.1. Small planet occurrence rates. There are a number of features of super-Earth oc-
currence rates that are relevant to possible shared origins channels (or lack thereof) with
hot Jupiters:
• Super-Earths are more common than giant planets. ∼40% of stars have at least
one super-Earth within a fifty day orbital period (Howard et al. 2010), whereas only
∼ 10% of stars have giant planets at any orbital period (e.g., Cumming et al. 2008,
Zechmeister et al. 2013).
• Super-Earths and Neptunes’ occurrence rate is constant beyond ten day orbital peri-
ods and drops within 10 days (see Lee & Chiang 2017 and references therein), whereas
the giant planet occurrence rate drops within ∼ 3 days (§3.4) and also exhibits a Val-
ley from ∼ 10–100 days (§4.3.1).
• The majority of super-Earths and Neptunes at sub year orbital periods are in multi-
planet systems (e.g., Lissauer, Dawson & Tremaine 2014, Ballard & Johnson 2016,
Dawson, Lee & Chiang 2016). However, hot Neptunes (2–6 Earth radii, P < 10 days),
recently termed Hoptunes, are most commonly in single transiting systems (Dong
et al. 2017). This distinction between warm Neptunes and Hoptunes is reminiscent of
that between warm Jupiters (accompanied by small nearby planets) vs. hot Jupiters
(rarely accompanied by small nearby planets).
• The occurrence rates and host-star metallicity dependence of Hoptunes is similar to
hot Jupiters (Dong et al. 2017); smaller planets at < 10 day orbital periods are more
common and less dependent on host star metallicity (e.g., Buchhave et al. 2012).
• The lack of Hoptunes orbiting M-dwarfs (e.g., Dressing & Charbonneau 2015) also
links them to hot Jupiters.
We now review how each of these features squares with different origins theories.
Under the in situ formation hypothesis, hot super-Earths form like hot Jupiters but fail
to achieve runaway gas accretion. Lee, Chiang & Ormel (2014) and Lee & Chiang (2016)
considered why super-Earths formed in situ may fail to grow into hot Jupiters. They argued
that if the local solid surface density is not high enough to form these bodies as isolation
masses (§2.1), they may need to wait to grow through giant impacts when the supply of
gas is nearly depleted. This explanation jives with the overall higher occurrence rate of
super-Earths (vs. giant planets) across all orbital periods. This explanation also accounts
for the metallicity dependence of gas giants (§4.1.3) and of hot Neptunes with low mass gas
envelopes (e.g., Dawson, Chiang & Lee 2015). However, no explanation has yet been posed
for super-Earths’ period cliff vs. giant planets’ period valley. Nor does in situ formation
satisfactorily explain the difference in nearby companions between hot vs. warm Jupiters.
Disk migration can deliver small planets to short orbital periods. Disk migration occurs
more quickly for super-Earths, which (except in very low viscosity disks) are not able to
open gaps that slow migration. Under the disk migration hypothesis, the overall higher
www.annualreviews.org • 39
occurrence rates of small planets and their metallicity dependence would reflect the efficiency
of their formation ex situ. However, disk migration is difficult to reconcile with super-Earths’
flat period distribution beyond 10 days (e.g., Lee & Chiang 2017). Early population studies
of planetary migration predicted a planet desert of warm super-Earth/Neptunes (e.g., Ida
& Lin 2008), at odds with the observed high occurrence rate.
A key challenge for accounting for hot super-Earths and Neptunes via high eccentricity
tidal migration is whether planetary systems have the requisite initial conditions to raise
the eccentricities of these small bodies. If outer systems of smaller planets resemble inner
ones, they are multi-planet systems of similar mass/size planets (Ciardi et al. 2013). Close
encounters among small planets lead to collisions rather than eccentricity excitation, even
out to several AU (eqn. 10), making planet-planet scattering ineffective at generating highly
elliptical orbits. Moreover, super-Earths in multi-planet systems keep each other safe from
the secular eccentricity excitation by binary stars and giant planets. Secular chaos is a
possible mechanism to generate large eccentricities but has not been explored in detail for
systems of small planets.
However, although disk migration and tidal migration seem unlikely to be making a
substantial contribution to the small planet population at large, they arguably play a role
in Hoptunes’ origins. Hoptunes are in danger of losing their atmospheres (e.g., Lopez,
Fortney & Miller 2012) while their stars are young and active. Young stars produce a
relatively large amount UV and X-ray flux, which can photo-evaporate low mass planets’
hydrogen and helium atmospheres. (Note: Hoptunes with close to 10 day orbital periods
could potentially retain their hydrogen and helium atmospheres if their cores are sufficiently
massive, e.g., Lopez & Fortney 2013. However, none of the ∼ 10 Hoptunes with mass
measurements has a sufficiently high mass or long orbital period.) For smaller Hoptunes
less than 4R⊕ (e.g., Lopez & Fortney 2014), formation ex situ could endow Hoptunes
with steam atmospheres that are less susceptible to photo-evaporation. Furthermore, high
eccentricity tidal migration can occur later in the system’s history, delivering Hoptunes
after the star’s ∼ 100 Myr active stage.
4.4.2. Stellar obliquities of small planet host stars. Measuring stellar obliquities of small
planets for comparison to hot Jupiters can potentially distinguish which physical processes
are predominantly shaping the stellar obliquitities of hot Jupiter hosts (§3.2; Fig. 5). If
spin-orbit misalignments are primarily caused by primordial misalignment of the disk (e.g.,
Batygin 2012), we would expect small planets and flat multi-planet systems to be misaligned
around hot stars as well. Like warm Jupiters (§4.3.4), small planets are not effective at
raising tides on their stars and we can study their obliquity distribution as one unsculpted
by tidal realignment. In a study using the five compact multi-planet systems at the time
with known obliquities, Albrecht et al. (2013) found that all are aligned with their star
and the collection is inconsistent with being drawn from an isotropic obliquity distribution.
However, all five stars had temperatures consistent with < 6250 K. Most recently, (Winn
et al. 2017a)) found that most host stars hosting planets have low obliquities. Of the six
possible misaligments they found, half were hot Jupiter hosts, whereas the vast majority
of the sample hosted small planets. This evidence supports the interpretation that hot
Jupiters were misaligned after the gas disk stage and counters the interpretation that hot
stars have high obliquitites independent of hot Jupiters. The TESS (Ricker et al. 2015),
CHEOPS (Broeg et al. 2013), and PLATO (Rauer et al. 2014) samples will provide more
opportunities to compare obliquities of hosts of hot Jupiters vs. small planets (e.g., Quinn
40
& White 2016).
4.4.3. Ultra-short Period Planets. Ultra-short period planets (USPs) are planets on sub-
day orbital periods. The overall occurrence rate of USPs is 0.5% for USPs greater than
0.84 R⊕(Sanchis-Ojeda et al. 2014); the majority of these USPs are Earths (0.84–1.25 R⊕),
rather than super-Earths. USPS tend reside in compact multi-planet systems (Sanchis-
Ojeda et al. 2014). Some may be tidally stripped remnants of hot Jupiters (e.g., Valsecchi
et al. 2015, Jackson et al. 2016). See Fig. 5 of Valsecchi et al. (2015) for examples of
planet trajectories in and out of the Roche limit as they evolve from hot Jupiters to super-
Earths. However, most USPs are likely not tidally stripped Jupiters because, unlike most
hot Jupiters, they reside in in compact multi-planet systems (Sanchis-Ojeda et al. 2014) and
they lack hot Jupiters’ host star metallicity dependence (Winn et al. 2017b). Lee & Chiang
(2017) showed that USPs can be accounted for by tidal evolution of compact multi-planet
systems. All these lines of evidence suggest that the majority of USPs may have a different
origin than hot Jupiters.
However, one observed feature does link USPs to giant planets: most of the special
systems, like WASP-47, containing a hot or warm giant planet in orbital resonance with
a neighbor also contain a USP (Fig. 8). Perhaps this special type of giant planet system
shares an origin with the more common compact, small multi-planet systems.
4.4.4. Summary: small planets and hot Jupiters. The links between small planets and hot
Jupiters are still uncertain. However, the following possibility seems consistent with all the
evidence. Perhaps the majority of super-Earths form in situ but some migration channel
(disk or tidal) delivers some hot Jupiters and some planets as well. Perhaps these small
planets generally blend in with the most abundant in situ super-Earth populations, except
the distinctive Hoptunes (Dong et al. 2017), whose gas envelopes are unlikely to survive if
Hoptunes formed in situ.
5. SUMMARY AND DISCUSSION
Despite thousands of observational and theoretical studies of hot Jupiters over the past
twenty years, we still have no consensus on the predominant channel for their origin. How-
ever, in this review, we have attempted to demonstrate that our community’s understanding
of their origin has advanced from speculation and post-dictions to detailed comparisons be-
tween observations and theories and to testable predictions for upcoming missions and
ongoing surveys. In §2, we described the three origins hypotheses for hot Jupiters: in
situ formation, disk migration, and tidal migration. In §3, we synthesized which observed
properties of hot Jupiters were consistent or inconsistent with these hypotheses. In §4, we
summarized how connections between hot Jupiters and other exoplanet populations provide
tests of hot Jupiters’ origins. We tabulated which hot Jupiter properties the three origins
hypotheses explain or fail to explain in Table 1.
Although no hypothesis for hot Jupiters’ origins can explain all the evidence, each
piece of evidence is explained by at least one origin channel. Throughout the review, we
have emphasized the power of two commonly operating origins channels to account for the
diversity of hot Jupiter properties. High eccentricity tidal migration triggered by planet-
planet Kozai-Lidov cycles is a strong contender for one of the two most prevalent origins
channels (§2.3). The supporting evidence is as follows. (We group related lines of evidence
www.annualreviews.org • 41
together, so the following list is not ordered by strength of evidence.)
• Tidal migration accounts for a lack small planets near most hot Jupiters (§4.2).
• Tidal migration accounts for the similarities in occurrence rates, multiplicity, and
stellar metallicity dependence between hot Jupiters and Hoptunes (§4.4).
• Moderately eccentric hot Jupiters cannot be explained by in situ formation or disk
migration; they are most consistent with tidal migration (§3.1).
• These eccentric hot Jupiters orbit metal-rich stars, implicating planet-planet inter-
actions in their origins (§3.1), rather than planet-stellar Kozai, because giant planet
formation is strongly correlated with stellar metallicity but stellar multiplicity is not.
• Most hot Jupiters have long period giant planet companions capable of driving Kozai-
Lidov cycles but few have stellar binary companions that are capable (§4.2).
• Planet-planet Kozai is the only mechanism shown to produce a sizable population of
eccentric warm Jupiters (§4.3). Moreover, these eccentric warm Jupiters orbit metal-
rich stars (§4.3) and are more likely to have outer giant planet companions (§4.2),
implicating planet-planet interactions.
We note that although we would summarize this type of dynamical history as tidal
migration triggered by planet-planet Kozai (e.g., Naoz et al. 2011, Nagasawa & Ida 2011,
Petrovich & Tremaine 2016; §2.3), other mechanisms may play important supporting roles.
For example, planet-planet scattering may establish the requisite mutual inclinations for
planet-planet Kozai, or disk migration may deliver the planets inside the ice line, if they
begin their journey inside where they formed. We also caution that this hypothesis is our
best attempt to synthesize the state of the field, rather than the universal consensus of
the community. A major open question is whether giant planet systems have the mutual
inclinations necessary for this mechanism to operate. Another open question is whether
proto-hot Jupiters had nearby small planets at their formation locations that would quench
secular interactions with more distant giant planets (though these may have been dislodged
during the planet-planet scattering process).
That idea that two origins channels are prevalent has early roots in the distribution of
hot Jupiter host star obliquities. Fabrycky & Winn (2009) and subsequent studies identified
two components to the obliquity distribution, one consisting of low obliquity (well-aligned
systems) and another component with a broad distribution. The interpretation of hot
Jupiter host star obliquities is currently ambiguous and no longer definitively linked to two
origins channels (§3.2,4.3.4, 4.4.2). Instead, we invoke a second origins channel – i.e., to
supplement tidal migration triggered by planet-planet Kozai – on following evidence:
• The eccentric hot Jupiters orbiting metal rich stars are in contrast to the mixture of
metal-rich and metal-poor stars hosting hot Jupiters (§3.1). This mixture of host star
metallicities for hot Jupiters implies another origins channel that produces circular
hot Jupiters orbiting lower metallicity stars.
• Two origins channels, correlated with host star metallicity, may contribute to the
Kepler’s sample lack of super-eccentric proto-hot Jupiters (i.e., generated from high
eccentricity tidal migration) and overall lack of hot Jupiters (§4.3.3) .
• The high eccentricity tidal migration scenarios are not sufficiently efficient at produc-
ing hot Jupiters at the observed rates (§4.1).
• High eccentricity tidal migration cannot easily account for circular warm Jupiters,
which are also correlated with lower host star metallicity (§4.3).
42
• High eccentricity tidal migration cannot account for warm Jupiters with nearby com-
panions (§4.3) or rare hot Jupiters with nearby companions, like WASP-47b (Becker
et al. 2015, §4.2).
• High eccentricity tidal migration cannot easily account for recent discoveries of hot
Jupiters orbiting T Tauri stars (§3.5).
• A number of hot Jupiters are observed on orbits within twice the Roche limit (§3.4),
where we don’t expect high-eccentricity tidal migration to deliver them. A second
channel could deliver hot Jupiters to 1 − −2aRoche whether requiring subsequent
evolution via tides raised on the star.
• Becker et al. (2017) recently found that detected planet companions in six hot Jupiters
systems cannot have sufficient mutual inclination to have driven Kozai-Lidov high
eccentricity tidal migration of the hot Jupiters.
The second origins channel could be either disk migration or in situ formation. For
assessing both theories, the dominant uncertainty is whether disk conditions are suitable to
produce the observed properties of hot Jupiters and their links to other planet populations.
In situ formation, not feasible through gravitational instability, requires a local huge disk
solid surface density for core accretion (§2.1). The solids likely need to be transported
from further out in the disk. Disk migration has been more successful in accounting for
the observed giant planet period distribution (§4.1), including the Period Valley (§4.3.1),
but requires tuned disk conditions. The inner limit of the hot Jupiter period distribution
is nominally more consistent with disk migration (§3.4), but subsequent tidal decay could
deliver planets formed in situ at the disk edge to shorter orbital periods (§3.4) for certain
stellar tidal parameters. Huang, Wu & Triaud (2016) argued that nearby small planet
companions of warm Jupiters are more consistent with in situ formation, but forming the
nearby planets in situ following disk migration may be a plausible alternative. Better
constraints on disk conditions and more detailed head-to-head comparisons between the
period distribution and warm Jupiters’ nearby companions would help distinguish whether
situ formation vs. disk migration is the predominant second channel.
Kozai oscillations driven by a wide stellar companion, followed by tidal circularization,
has been extensively investigated in the origins of hot Jupiters (e.g., Wu & Murray 2003,
Fabrycky & Tremaine 2007). However, our interpretation of the current state of the field
is that it cannot be a predominant channel, though it could still be operating in a handful
of individual systems. Consistent with this mechanism not playing a major role in shaping
the architectures of planetary systems, Ngo et al. (2017) found no significant differences in
the mass, orbital eccentricity, and semi-major axis distribution of the innermost planet in
the multi-stellar systems vs. single-star systems.
To resolve the outstanding issues of hot Jupiters’ origins, we recommend follow-up
observational and theoretical studies related to the following issues:
FUTURE ISSUES
1. Eccentric hot Jupiters: several lines of evidence hinge on these planets, motivating
continued, intensive radial-velocity follow up to map out the architecture of their
systems. Do they all have giant planet companions that were capable of triggering
their high eccentricity migration? And if so, were they protected from its influence
by smaller planets in the system? Can we rule out for each eccentric hot Jupiter
www.annualreviews.org • 43
that a nearby eccentric giant planet excited their eccentricity in situ?
2. Many transiting hot Jupiters have limited radial-velocity follow-up designed only
to confirm them as planets, not to constrain their eccentricities. Continued radial-
velocity follow-up – e.g., as pursued by the HARPS GAPS program (e.g., Bonomo
et al. 2017) – could increase the sample of valuable eccentric warm Jupiters.
3. For hot Jupiters with giant planet companions, Gaia astrometry may diagnose
whether the giant planet companion has the necessary mutual inclination to trigger
high eccentricity migration using Kozai-Lidov cycles (Casertano et al. 2008). We
recommend similar astrometric follow-up for companions of eccentric warm Jupiters.
4. Constraining the mutual inclination distribution of giant planets across all orbital
periods will address the applicability of planet-planet Kozai. In addition to Gaia
astrometry (Casertano et al. 2008), constraints from transit timing and duration
variations of transiting Jupiters can contribute.
5. Theoretical and observational investigations should thoroughly constrain how often
intervening small planets interfere with secular interactions between giant planets.
6. The biggest outstanding question for in situ formation is whether the requisite
disk conditions are seen in nature. We can address this question with further
observational and theoretical studies of solid transport within gas disks.
7. More theoretical studies are needed to address whether hot Jupiters’ atmospheric
properties can feasibly trace formation location or whether the diversity of disk
conditions make this infeasible even for a large sample.
8. A larger sample of obliquitities of warm Jupiters hosts and small planets hosts will
inform us of the extent to which the primordial distribution of hot Jupiter hosts
has been altered by tides.
9. To harness the power of host star ages to distinguish among origins hypotheses
(§3.5), we need a large sample of stellar ages, ideally extending to very young ages.
Gaia (Gaia Collaboration et al. 2016), TESS (Campante et al. 2016), and PLATO
(Rauer et al. 2014) will enlarge the sample of stellar age estimates, and ongoing
surveys of T Tauri stars (e.g., Yu & MaTYSSE Collaboration 2017) can assess how
often hot Jupiters achieve their short orbital periods during the gas disk stage.
6. DISCLOSURE STATEMENT
The authors are not aware of any affiliations, sponsorships, funding, or financial holdings
that might be perceived as affecting the objectivity of this review.
7. ACKNOWLEDGMENTS
We thank Eric Ford, Sivan Ginzburg, Chelsea Huang, Taisiya Kopytova, David Latham,
Eve Lee, Gongjie Li, Henry Ngo, Ana-Maria Piso, Johanna Teske, and Yanqin Wu for help-
ful conversations. We are grateful to Simon Albrecht, Thomas Beatty, Chelsea Huang, Eve
Lee, Smadar Naoz, Cristobal Petrovich, and Christopher Spalding for insightful comments
on a manuscript draft. We thank Ewine van Dishoeck and two anonymous reviewers for
helpful comments that improved the review. We thank Jose Manuel Almenara, Francesca
Faedi, Michael Gillon, Guillaume Hebrard, and Rachel Street for providing us with up-
44
per limits on the eccentricities of certain hot Jupiters based on their published fits. This
work was partially supported by funding from the NASA Exoplanets Research program
(NNX16AB50G). The Center for Exoplanets and Habitable Worlds is supported by the
Pennsylvania State University, the Eberly College of Science, and the Pennsylvania Space
Grant Consortium.
LITERATURE CITED
Albrecht S, Winn JN, Johnson JA, Howard AW, Marcy GW, et al. 2012. Ap. J. 757:18
Albrecht S, Winn JN, Marcy GW, Howard AW, Isaacson H, Johnson JA. 2013. Ap. J. 771:11
Alexander RD, Pascucci I. 2012. MNRAS 422:L82–L86
Alibert Y, Mordasini C, Benz W, Winisdoerffer C. 2005. Astron. Astrophys. 434:343–353
Anderson KR, Lai D. 2017. ArXiv e-prints arXiv:1706.00084
Anderson KR, Storch NI, Lai D. 2016. MNRAS 456:3671–3701
Andrews SM. 2015. Pub. Astron. Soc. Pac. 127:961–993
Ansdell M, Williams JP, Manara CF, Miotello A, Facchini S, et al. 2017. Astron. J. 153:240
Ansdell M, Williams JP, van der Marel N, Carpenter JM, Guidi G, et al. 2016. Ap. J. 828:46
Antognini JMO. 2015. MNRAS 452:3610–3619
Antonini F, Hamers AS, Lithwick Y. 2016. Astron. J. 152:174
Armitage PJ. 2013. Astrophysics of Planet Formation
Arras P, Socrates A. 2010. Ap. J. 714:1–12
Ballard S, Johnson JA. 2016. Ap. J. 816:66
Barenfeld SA, Carpenter JM, Ricci L, Isella A. 2016. Ap. J. 827:142
Barker AJ. 2016. MNRAS 460:2339–2350
Barnes JW. 2009. Ap. J. 705:683–692
Baruteau C, Crida A, Paardekooper SJ, Masset F, Guilet J, et al. 2014. Protostars and Planets VI
:667–689
Batygin K. 2012. Nature 491:418–420
Batygin K, Bodenheimer PH, Laughlin GP. 2016. Ap. J. 829:114
Batygin K, Stevenson DJ. 2010. Ap. J. Let. 714:L238–L243
Beauge C, Nesvorny D. 2012. Ap. J. 751:119
Becker JC, Vanderburg A, Adams FC, Khain T, Bryan M. 2017. Astron. J. 154:230
Becker JC, Vanderburg A, Adams FC, Rappaport SA, Schwengeler HM. 2015. Ap. J. Let. 812:L18
Bitsch B, Crida A, Libert AS, Lega E. 2013. Astron. Astrophys. 555:A124
Bodenheimer P, Lin DNC, Mardling RA. 2001. Ap. J. 548:466–472
Boley AC, Granados Contreras AP, Gladman B. 2016. Ap. J. Let. 817:L17
Bonomo AS, Desidera S, Benatti S, Borsa F, Crespi S, et al. 2017. Astron. Astrophys. 602:A107
Boss AP. 1997. Science 276:1836–1839
Broeg C, Fortier A, Ehrenreich D, Alibert Y, Baumjohann W, et al. 2013. In European Physical
Journal Web of Conferences, vol. 47 of European Physical Journal Web of Conferences
Brucalassi A, Koppenhoefer J, Saglia R, Pasquini L, Ruiz MT, et al. 2017. Astron. Astrophys.
603:A85
Bryan ML, Knutson HA, Howard AW, Ngo H, Batygin K, et al. 2016. Ap. J. 821:89
Buchhave LA, Latham DW, Johansen A, Bizzarro M, Torres G, et al. 2012. Nature 486:375–377
Buhler PB, Knutson HA, Batygin K, Fulton BJ, Fortney JJ, et al. 2016. Ap. J. 821:26
Campante TL, Schofield M, Kuszlewicz JS, Bouma L, Chaplin WJ, et al. 2016. Ap. J. 830:138
Casertano S, Lattanzi MG, Sozzetti A, Spagna A, Jancart S, et al. 2008. Astron. Astrophys.
482:699–729
Catala C, Donati JF, Shkolnik E, Bohlender D, Alecian E. 2007. MNRAS 374:L42–L46
Chabrier G, Johansen A, Janson M, Rafikov R. 2014. Protostars and Planets VI :619–642
Chambers JE. 2016. Ap. J. 825:63
www.annualreviews.org • 45
Chambers JE, Wetherill GW, Boss AP. 1996. Icarus 119:261–268
Chang SH, Gu PG, Bodenheimer PH. 2010. Ap. J. 708:1692–1702
Charbonneau D, Brown TM, Latham DW, Mayor M. 2000. Ap. J. Let. 529:L45–L48
Chatterjee S, Ford EB, Geller AM, Rasio FA. 2012. MNRAS 427:1587–1602
Chatterjee S, Ford EB, Matsumura S, Rasio FA. 2008. Ap. J. 686:580–602
Chiang E, Laughlin G. 2013. MNRAS 431:3444–3455
Ciardi DR, Fabrycky DC, Ford EB, Gautier III TN, Howell SB, et al. 2013. Ap. J. 763:41
Coleman GAL, Nelson RP. 2016. MNRAS 460:2779–2795
Collier Cameron A, Guenther E, Smalley B, McDonald I, Hebb L, et al. 2010. MNRAS 407:507–514
Cumming A, Butler RP, Marcy GW, Vogt SS, Wright JT, Fischer DA. 2008. Pub. Astron. Soc.
Pac. 120:531
Damiani C, Dıaz RF. 2016. Astron. Astrophys. 589:A55
Damiani C, Lanza AF. 2015. Astron. Astrophys. 574:A39
Dawson RI. 2014. Ap. J. Let. 790:L31
Dawson RI, Chiang E. 2014. Science 346:212–216
Dawson RI, Chiang E, Lee EJ. 2015. MNRAS 453:1471–1483
Dawson RI, Johnson JA. 2012. Ap. J. 756:122
Dawson RI, Johnson JA, Fabrycky DC, Foreman-Mackey D, Murray-Clay RA, et al. 2014. Ap. J.
791:89
Dawson RI, Johnson JA, Morton TD, Crepp JR, Fabrycky DC, et al. 2012. Ap. J. 761:163
Dawson RI, Lee EJ, Chiang E. 2016. Ap. J. 822:54
Dawson RI, Murray-Clay RA. 2013. Ap. J. Let. 767:L24
Dawson RI, Murray-Clay RA, Johnson JA. 2015. Ap. J. 798:66
Donati JF, Moutou C, Malo L, Baruteau C, Yu L, et al. 2016. Nature 534:662–666
Dong R, Dawson R. 2016. Ap. J. 825:77
Dong S, Katz B, Socrates A. 2014. Ap. J. Let. 781:L5
Dong S, Xie JW, Zhou JL, Zheng Z, Luo A. 2017. ArXiv e-prints arXiv:1706.07807
Dressing CD, Charbonneau D. 2015. Ap. J. 807:45
Duffell PC, Chiang E. 2015. Ap. J. 812:94
Duffell PC, Haiman Z, MacFadyen AI, D’Orazio DJ, Farris BD. 2014. Ap. J. Let. 792:L10
Durisen RH, Boss AP, Mayer L, Nelson AF, Quinn T, Rice WKM. 2007. Protostars and Planets
V :607–622
Durmann C, Kley W. 2015. Astron. Astrophys. 574:A52
Eggleton PP, Kiseleva LG, Hut P. 1998. Ap. J. 499:853–870
Eisner JA, Hillenbrand LA, White RJ, Akeson RL, Sargent AI. 2005. Ap. J. 623:952–966
Endl M, Caldwell DA, Barclay T, Huber D, Isaacson H, et al. 2014. Ap. J. 795:151
Evans DF, Southworth J, Maxted PFL, Skottfelt J, Hundertmark M, et al. 2016. Astron. Astrophys.
589:A58
Faber JA, Rasio FA, Willems B. 2005. Icarus 175:248–262
Fabrycky D, Tremaine S. 2007. Ap. J. 669:1298–1315
Fabrycky DC. 2010. Non-Keplerian Dynamics of Exoplanets. 217–238
Fabrycky DC, Winn JN. 2009. Ap. J. 696:1230–1240
Fedele D, van den Ancker ME, Henning T, Jayawardhana R, Oliveira JM. 2010. Astron. Astrophys.
510:A72
Fielding DB, McKee CF, Socrates A, Cunningham AJ, Klein RI. 2015. MNRAS 450:3306–3318
Fischer DA, Valenti J. 2005. Ap. J. 622:1102–1117
Ford EB, Rasio FA. 2006. Ap. J. Let. 638:L45–L48
Gaia Collaboration, Prusti T, de Bruijne JHJ, Brown AGA, Vallenari A, et al. 2016. Astron.
Astrophys. 595:A1
Gaspar A, Rieke GH, Ballering N. 2016. Ap. J. 826:171
Gaudi BS, Seager S, Mallen-Ornelas G. 2005. Ap. J. 623:472–481
46
Ginzburg S, Sari R. 2016. Ap. J. 819:116
Goldreich P, Lithwick Y, Sari R. 2004. Annu. Rev. Astron. Astrophy. 42:549–601
Goldreich P, Sari R. 2003. Ap. J. 585:1024–1037
Goldreich P, Schlichting HE. 2014. Astron. J. 147:32
Goldreich P, Tremaine S. 1980. Ap. J. 241:425–441
Gonzalez G. 1997. MNRAS 285:403–412
Greenzweig Y, Lissauer JJ. 1990. Icarus 87:40–77
Guillochon J, Ramirez-Ruiz E, Lin D. 2011. Ap. J. 732:74
Guillot T. 2005. Annual Review of Earth and Planetary Sciences 33:493–530
Guillot T, Showman AP. 2002. Astron. Astrophys. 385:156–165
Guo X, Johnson JA, Mann AW, Kraus AL, Curtis JL, Latham DW. 2017. Ap. J. 838:25
Hamers AS, Antonini F, Lithwick Y, Perets HB, Portegies Zwart SF. 2017. MNRAS 464:688–701
Hansen BMS. 2010. Ap. J. 723:285–299
Hansen BMS, Murray N. 2013. Ap. J. 775:53
Hao W, Kouwenhoven MBN, Spurzem R. 2013. MNRAS 433:867–877
Hartman JD, Bakos GA, Bhatti W, Penev K, Bieryla A, et al. 2016. Astron. J. 152:182
Hayashi C. 1981. Progress of Theoretical Physics Supplement 70:35–53
Heng K, Lyons JR. 2016. Ap. J. 817:149
Henry GW, Marcy GW, Butler RP, Vogt SS. 2000. Ap. J. Let. 529:L41–L44
Howard AW, Marcy GW, Johnson JA, Fischer DA, Wright JT, et al. 2010. Science 330:653
Huang C, Wu Y, Triaud AHMJ. 2016. Ap. J. 825:98
Huber D, Carter JA, Barbieri M, Miglio A, Deck KM, et al. 2013. Science 342:331–334
Hut P. 1981. Astron. Astrophys. 99:126–140
Ibgui L, Spiegel DS, Burrows A. 2011. Ap. J. 727:75
Ida S, Lin DNC. 2008. Ap. J. 673:487–501
Ida S, Lin DNC, Nagasawa M. 2013. Ap. J. 775:42
Jackson B, Jensen E, Peacock S, Arras P, Penev K. 2016. Celestial Mechanics and Dynamical
Astronomy 126:227–248
Jenkins JS, Jones HRA, Tuomi M, Dıaz M, Cordero JP, et al. 2017. MNRAS 466:443–473
Johansen A, Lambrechts M. 2017. Annual Review of Earth and Planetary Sciences 45:359–387
Johnson JA, Aller KM, Howard AW, Crepp JR. 2010. Pub. Astron. Soc. Pac. 122:905–915
Jones HRA, Butler RP, Tinney CG, Marcy GW, Penny AJ, et al. 2003. MNRAS 341:948–952
Juric M, Tremaine S. 2008. Ap. J. 686:603–620
Katz B, Dong S, Malhotra R. 2011. Physical Review Letters 107:181101
Knutson HA, Fulton BJ, Montet BT, Kao M, Ngo H, et al. 2014. Ap. J. 785:126
Kovacs G, Hartman JD, Bakos GA, Quinn SN, Penev K, et al. 2014. MNRAS 442:2081–2093
Kozai Y. 1962. Astron. J. 67:591
Kraft RP. 1967. Ap. J. 150:551
Kreidberg L, Line MR, Bean JL, Stevenson KB, Desert JM, et al. 2015. Ap. J. 814:66
Kuchner MJ, Lecar M. 2002. Ap. J. Let. 574:L87–L89
Lai D. 2012. MNRAS 423:486–492
Laskar J. 2008. Icarus 196:1–15
Latham DW, Rowe JF, Quinn SN, Batalha NM, Borucki WJ, et al. 2011. Ap. J. Let. 732:L24
Lecar M, Podolak M, Sasselov D, Chiang E. 2006. Ap. J. 640:1115–1118
Leconte J, Chabrier G, Baraffe I, Levrard B. 2010. Astron. Astrophys. 516:A64
Lee EJ, Chiang E. 2016. Ap. J. 817:90
Lee EJ, Chiang E. 2017. Ap. J. 842:40
Lee EJ, Chiang E, Ormel CW. 2014. Ap. J. 797:95
Lee MH, Peale SJ. 2002. Ap. J. 567:596–609
Li G, Naoz S, Holman M, Loeb A. 2014a. Ap. J. 791:86
Li G, Naoz S, Kocsis B, Loeb A. 2014b. Ap. J. 785:116
www.annualreviews.org • 47
Li G, Naoz S, Valsecchi F, Johnson JA, Rasio FA. 2014c. Ap. J. 794:131
Li G, Winn JN. 2016. Ap. J. 818:5
Lidov ML. 1962. Planet. Space Sci. 9:719–759
Lin DNC, Bodenheimer P, Richardson DC. 1996. Nature 380:606–607
Lin DNC, Papaloizou J. 1986. Ap. J. 309:846–857
Lin Y, Ogilvie GI. 2017. MNRAS 468:1387–1397
Lissauer JJ, Dawson RI, Tremaine S. 2014. Nature 513:336–344
Lopez ED, Fortney JJ. 2013. Ap. J. 776:2
Lopez ED, Fortney JJ. 2014. Ap. J. 792:1
Lopez ED, Fortney JJ. 2016. Ap. J. 818:4
Lopez ED, Fortney JJ, Miller N. 2012. Ap. J. 761:59
Madhusudhan N, Amin MA, Kennedy GM. 2014. Ap. J. Let. 794:L12
Malhotra R. 1993. Nature 365:819–821
Marcy GW, Butler RP. 2000. Pub. Astron. Soc. Pac. 112:137–140
Masset FS, Papaloizou JCB. 2003. Ap. J. 588:494–508
Masuda K. 2017. Astron. J. 154:64
Matsakos T, Konigl A. 2015. Ap. J. Let. 809:L20
Matsumura S, Peale SJ, Rasio FA. 2010. Ap. J. 725:1995–2016
Mayor M, Marmier M, Lovis C, Udry S, Segransan D, et al. 2011. ArXiv e-prints arXiv:1109.2497
Mayor M, Queloz D. 1995. Nature 378:355–359
Mazeh T, Perets HB, McQuillan A, Goldstein ES. 2015. Ap. J. 801:3
McLaughlin DB. 1924. Ap. J.
Miller BP, Gallo E, Wright JT, Pearson EG. 2015. Ap. J. 799:163
Millholland S, Wang S, Laughlin G. 2016. Ap. J. Let. 823:L7
Minton DA, Malhotra R. 2011. Ap. J. 732:53
Moro-Martın A, Marshall JP, Kennedy G, Sibthorpe B, Matthews BC, et al. 2015. Ap. J. 801:143
Morton TD, Johnson JA. 2011. Ap. J. 729:138
Munoz DJ, Lai D, Liu B. 2016. MNRAS 460:1086–1093
Mustill AJ, Davies MB, Johansen A. 2015. Ap. J. 808:14
Mustill AJ, Davies MB, Johansen A. 2017. MNRAS 468:3000–3023
Naef D, Latham DW, Mayor M, Mazeh T, Beuzit JL, et al. 2001. Astron. Astrophys. 375:L27–L30
Nagasawa M, Ida S. 2011. Ap. J. 742:72
Naoz S. 2016. Annu. Rev. Astron. Astrophy. 54:441–489
Naoz S, Farr WM, Lithwick Y, Rasio FA, Teyssandier J. 2011. Nature 473:187–189
Naoz S, Farr WM, Rasio FA. 2012. Ap. J. Let. 754:L36
Nelson BE, Ford EB, Rasio FA. 2017. Astron. J. 154:106
Neveu-VanMalle M, Queloz D, Anderson DR, Brown DJA, Collier Cameron A, et al. 2016. Astron.
Astrophys. 586:A93
Ngo H, Knutson HA, Bryan ML, Blunt S, Nielsen EL, et al. 2017. Astron. J. 153:242
Ngo H, Knutson HA, Hinkley S, Bryan M, Crepp JR, et al. 2016. Ap. J. 827:8
Oberg KI, Murray-Clay R, Bergin EA. 2011. Ap. J. Let. 743:L16
Obermeier C, Koppenhoefer J, Saglia RP, Henning T, Bender R, et al. 2016. Astron. Astrophys.
587:A49
Ogihara M, Inutsuka Si, Kobayashi H. 2013. Ap. J. Let. 778:L9
Ogihara M, Kobayashi H, Inutsuka Si. 2014. Ap. J. 787:172
Otor OJ, Montet BT, Johnson JA, Charbonneau D, Collier-Cameron A, et al. 2016. Astron. J.
152:165
Paardekooper SJ, Mellema G. 2006. Astron. Astrophys. 459:L17–L20
Pascucci I, Testi L, Herczeg GJ, Long F, Manara CF, et al. 2016. Ap. J. 831:125
Perri F, Cameron AGW. 1974. Icarus 22:416–425
Petrovich C. 2015a. Ap. J. 805:75
48
Petrovich C. 2015b. Ap. J. 799:27
Petrovich C, Tremaine S. 2016. Ap. J. 829:132
Petrovich C, Tremaine S, Rafikov R. 2014. Ap. J. 786:101
Piskorz D, Knutson HA, Ngo H, Muirhead PS, Batygin K, et al. 2015. Ap. J. 814:148
Piso AMA, Oberg KI, Birnstiel T, Murray-Clay RA. 2015. Ap. J. 815:109
Piso AMA, Youdin AN, Murray-Clay RA. 2015. Ap. J. 800:82
Plavchan P, Bilinski C. 2013. Ap. J. 769:86
Pollack JB, Hubickyj O, Bodenheimer P, Lissauer JJ, Podolak M, Greenzweig Y. 1996. Icarus
124:62–85
Pont F, Husnoo N, Mazeh T, Fabrycky D. 2011. MNRAS 414:1278–1284
Poppenhaeger K, Wolk SJ. 2014. Astron. Astrophys. 565:L1
Quinn SN, White RJ. 2016. Ap. J. 833:173
Quinn SN, White RJ, Latham DW, Buchhave LA, Cantrell JR, et al. 2012. Ap. J. Let. 756:L33
Quinn SN, White RJ, Latham DW, Buchhave LA, Torres G, et al. 2014. Ap. J. 787:27
Rafikov RR. 2004. Astron. J. 128:1348–1363
Rafikov RR. 2005. Ap. J. Let. 621:L69–L72
Rafikov RR. 2006. Ap. J. 648:666–682
Rasio FA, Ford EB. 1996. Science 274:954–956
Rauer H, Catala C, Aerts C, Appourchaux T, Benz W, et al. 2014. Experimental Astronomy
38:249–330
Raymond SN, Mandell AM, Sigurdsson S. 2006. Science 313:1413–1416
Reffert S, Bergmann C, Quirrenbach A, Trifonov T, Kunstler A. 2015. Astron. Astrophys. 574:A116
Rice WKM, Armitage PJ, Hogg DF. 2008. MNRAS 384:1242–1248
Ricker GR, Winn JN, Vanderspek R, Latham DW, Bakos GA, et al. 2015. Journal of Astronomical
Telescopes, Instruments, and Systems 1:014003
Rogers TM, Lin DNC, Lau HHB. 2012. Ap. J. Let. 758:L6
Rossiter RA. 1924. Ap. J.
Sanchis-Ojeda R, Rappaport S, Winn JN, Kotson MC, Levine A, El Mellah I. 2014. Ap. J. 787:47
Sanchis-Ojeda R, Winn JN, Holman MJ, Carter JA, Osip DJ, Fuentes CI. 2011. Ap. J. 733:127
Santerne A, Moutou C, Tsantaki M, Bouchy F, Hebrard G, et al. 2016. Astron. Astrophys. 587:A64
Santos NC, Bouchy F, Mayor M, Pepe F, Queloz D, et al. 2004. Astron. Astrophys. 426:L19–L23
Santos NC, Israelian G, Mayor M. 2001. Astron. Astrophys. 373:1019–1031
Santos NC, Israelian G, Mayor M. 2004. Astron. Astrophys. 415:1153–1166
Santos NC, Israelian G, Mayor M, Rebolo R, Udry S. 2003. Astron. Astrophys. 398:363–376
Schlaufman KC. 2010. Ap. J. 719:602–611
Schlaufman KC, Winn JN. 2016. Ap. J. 825:62
Schlichting HE. 2014. Ap. J. Let. 795:L15
Schneider J, Dedieu C, Le Sidaner P, Savalle R, Zolotukhin I. 2011. Astron. Astrophys. 532:A79
Shabram M, Demory BO, Cisewski J, Ford EB, Rogers L. 2016. Ap. J. 820:93
Shara MM, Hurley JR, Mardling RA. 2016. Ap. J. 816:59
Sing DK, Fortney JJ, Nikolov N, Wakeford HR, Kataria T, et al. 2016. Nature 529:59–62
Socrates A. 2013. ArXiv e-prints arXiv:1304.4121
Socrates A, Katz B, Dong S, Tremaine S. 2012. Ap. J. 750:106
Sousa SG, Santos NC, Israelian G, Mayor M, Udry S. 2011. Astron. Astrophys. 533:A141
Spalding C, Batygin K. 2015. Ap. J. 811:82
Spiegel DS, Burrows A. 2012. Ap. J. 745:174
Steffen JH, Ragozzine D, Fabrycky DC, Carter JA, Ford EB, et al. 2012. Proceedings of the National
Academy of Science 109:7982–7987
Stevenson DJ. 1982. Planet. Space Sci. 30:755–764
Storch NI, Anderson KR, Lai D. 2014. Science 345:1317–1321
Struve O. 1952. The Observatory 72:199–200
www.annualreviews.org • 49
Teyssandier J, Naoz S, Lizarraga I, Rasio FA. 2013. Ap. J. 779:166
Thommes EW, Duncan MJ, Levison HF. 1999. Nature 402:635–638
Toomre A. 1964. Ap. J. 139:1217–1238
Triaud AHMJ. 2011. Astron. Astrophys. 534:L6
Triaud AHMJ. 2017. ArXiv e-prints arXiv:1709.06376
Trilling DE, Benz W, Guillot T, Lunine JI, Hubbard WB, Burrows A. 1998. Ap. J. 500:428–439
Tsang D, Turner NJ, Cumming A. 2014. Ap. J. 782:113
Udry S, Mayor M, Santos NC. 2003. Astron. Astrophys. 407:369–376
Valsecchi F, Rappaport S, Rasio FA, Marchant P, Rogers LA. 2015. Ap. J. 813:101
Valsecchi F, Rasio FA, Steffen JH. 2014. Ap. J. Let. 793:L3
Wahl SM, Hubbard WB, Militzer B, Guillot T, Miguel Y, et al. 2017. GeoRL 44:4649–4659
Ward WR. 1997. Icarus 126:261–281
Weidenschilling SJ, Marzari F. 1996. Nature 384:619–621
Weiss LM, Marcy GW, Rowe JF, Howard AW, Isaacson H, et al. 2013. Ap. J. 768:14
Winn JN, Fabrycky D, Albrecht S, Johnson JA. 2010. Ap. J. Let. 718:L145–L149
Winn JN, Petigura EA, Morton TD, Weiss LM, Dai F, et al. 2017a. Astron. J. 154:270
Winn JN, Sanchis-Ojeda R, Rogers L, Petigura EA, Howard AW, et al. 2017b. Astron. J. 154:60
Wittenmyer RA, O’Toole SJ, Jones HRA, Tinney CG, Butler RP, et al. 2010. Ap. J. 722:1854–1863
Wright JT, Marcy GW, Howard AW, Johnson JA, Morton TD, Fischer DA. 2012. Ap. J. 753:160
Wu Y. 2017. ArXiv e-prints arXiv:1710.02542
Wu Y, Lithwick Y. 2011. Ap. J. 735:109
Wu Y, Lithwick Y. 2013. Ap. J. 763:13
Wu Y, Murray N. 2003. Ap. J. 589:605–614
Wu Y, Murray NW, Ramsahai JM. 2007. Ap. J. 670:820–825
Xiang-Gruess M. 2016. MNRAS 455:3086–3100
Xu W, Lai D. 2016. MNRAS 459:2925–2939
Youdin AN, Mitchell JL. 2010. Ap. J. 721:1113–1126
Yu L, Donati JF, Hebrard EM, Moutou C, Malo L, et al. 2017. MNRAS 467:1342–1359
Yu L, MaTYSSE Collaboration. 2017. In Living Around Active Stars, eds. D Nandy, A Valio,
P Petit, vol. 328 of IAU Symposium
Zechmeister M, Kurster M, Endl M, Lo Curto G, Hartman H, et al. 2013. Astron. Astrophys.
552:A78
50